WO2024047608A1 - Procédés et systèmes de profilage immunitaire basé sur l'ia pour la stratification de patients cancéreux - Google Patents
Procédés et systèmes de profilage immunitaire basé sur l'ia pour la stratification de patients cancéreux Download PDFInfo
- Publication number
- WO2024047608A1 WO2024047608A1 PCT/IB2023/058673 IB2023058673W WO2024047608A1 WO 2024047608 A1 WO2024047608 A1 WO 2024047608A1 IB 2023058673 W IB2023058673 W IB 2023058673W WO 2024047608 A1 WO2024047608 A1 WO 2024047608A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- til
- patient
- expression level
- features
- neural network
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 214
- 206010028980 Neoplasm Diseases 0.000 title claims description 210
- 201000011510 cancer Diseases 0.000 title claims description 44
- 238000011493 immune profiling Methods 0.000 title description 2
- 238000013517 stratification Methods 0.000 title description 2
- 210000003171 tumor-infiltrating lymphocyte Anatomy 0.000 claims abstract description 456
- 102000008096 B7-H1 Antigen Human genes 0.000 claims abstract description 227
- 108010074708 B7-H1 Antigen Proteins 0.000 claims abstract description 227
- 101100519207 Mus musculus Pdcd1 gene Proteins 0.000 claims abstract description 201
- 230000004044 response Effects 0.000 claims abstract description 169
- 238000003062 neural network model Methods 0.000 claims abstract description 162
- 230000001225 therapeutic effect Effects 0.000 claims abstract description 161
- 238000002619 cancer immunotherapy Methods 0.000 claims abstract description 138
- 238000004422 calculation algorithm Methods 0.000 claims description 158
- 210000004027 cell Anatomy 0.000 claims description 119
- 230000011218 segmentation Effects 0.000 claims description 118
- 238000007635 classification algorithm Methods 0.000 claims description 73
- 238000013528 artificial neural network Methods 0.000 claims description 53
- 210000004881 tumor cell Anatomy 0.000 claims description 44
- 206010005003 Bladder cancer Diseases 0.000 claims description 30
- 230000035772 mutation Effects 0.000 claims description 30
- 208000007097 Urinary Bladder Neoplasms Diseases 0.000 claims description 28
- 201000005112 urinary bladder cancer Diseases 0.000 claims description 28
- 238000011467 adoptive cell therapy Methods 0.000 claims description 26
- 229940076838 Immune checkpoint inhibitor Drugs 0.000 claims description 23
- 239000012274 immune-checkpoint protein inhibitor Substances 0.000 claims description 23
- 101150067539 AMBP gene Proteins 0.000 claims description 20
- 102000006306 Antigen Receptors Human genes 0.000 claims description 20
- 108010083359 Antigen Receptors Proteins 0.000 claims description 20
- 210000001744 T-lymphocyte Anatomy 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 13
- 238000010801 machine learning Methods 0.000 claims description 12
- 239000012270 PD-1 inhibitor Substances 0.000 claims description 10
- 239000012668 PD-1-inhibitor Substances 0.000 claims description 10
- 239000012271 PD-L1 inhibitor Substances 0.000 claims description 10
- 210000002865 immune cell Anatomy 0.000 claims description 10
- 229940121655 pd-1 inhibitor Drugs 0.000 claims description 10
- 229940121656 pd-l1 inhibitor Drugs 0.000 claims description 10
- 206010006187 Breast cancer Diseases 0.000 claims description 8
- 208000026310 Breast neoplasm Diseases 0.000 claims description 8
- 108010019670 Chimeric Antigen Receptors Proteins 0.000 claims description 8
- 208000017604 Hodgkin disease Diseases 0.000 claims description 8
- 208000021519 Hodgkin lymphoma Diseases 0.000 claims description 8
- 208000010747 Hodgkins lymphoma Diseases 0.000 claims description 8
- 208000008839 Kidney Neoplasms Diseases 0.000 claims description 8
- 206010038389 Renal cancer Diseases 0.000 claims description 8
- 108091008874 T cell receptors Proteins 0.000 claims description 8
- 102000016266 T-Cell Antigen Receptors Human genes 0.000 claims description 8
- 201000010982 kidney cancer Diseases 0.000 claims description 8
- 201000001441 melanoma Diseases 0.000 claims description 8
- 210000000581 natural killer T-cell Anatomy 0.000 claims description 8
- 210000000822 natural killer cell Anatomy 0.000 claims description 8
- 208000002154 non-small cell lung carcinoma Diseases 0.000 claims description 8
- 208000029729 tumor suppressor gene on chromosome 11 Diseases 0.000 claims description 8
- 239000002955 immunomodulating agent Substances 0.000 claims description 5
- 229940121354 immunomodulator Drugs 0.000 claims description 5
- 210000001519 tissue Anatomy 0.000 description 153
- 238000011282 treatment Methods 0.000 description 30
- 238000004458 analytical method Methods 0.000 description 22
- 239000003814 drug Substances 0.000 description 20
- 239000000203 mixture Substances 0.000 description 9
- 230000017074 necrotic cell death Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 238000003860 storage Methods 0.000 description 7
- 238000002560 therapeutic procedure Methods 0.000 description 7
- 210000003855 cell nucleus Anatomy 0.000 description 6
- 238000004891 communication Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 108010077544 Chromatin Proteins 0.000 description 5
- 210000003483 chromatin Anatomy 0.000 description 5
- 201000010099 disease Diseases 0.000 description 5
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 5
- WZUVPPKBWHMQCE-UHFFFAOYSA-N Haematoxylin Chemical compound C12=CC(O)=C(O)C=C2CC2(O)C1C1=CC=C(O)C(O)=C1OC2 WZUVPPKBWHMQCE-UHFFFAOYSA-N 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000036541 health Effects 0.000 description 4
- 238000009169 immunotherapy Methods 0.000 description 4
- 238000011275 oncology therapy Methods 0.000 description 4
- 239000000090 biomarker Substances 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000007481 next generation sequencing Methods 0.000 description 3
- 208000024891 symptom Diseases 0.000 description 3
- 102000037982 Immune checkpoint proteins Human genes 0.000 description 2
- 108091008036 Immune checkpoint proteins Proteins 0.000 description 2
- 238000001574 biopsy Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- YQGOJNYOYNNSMM-UHFFFAOYSA-N eosin Chemical compound [Na+].OC(=O)C1=CC=CC=C1C1=C2C=C(Br)C(=O)C(Br)=C2OC2=C(Br)C(O)=C(Br)C=C21 YQGOJNYOYNNSMM-UHFFFAOYSA-N 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000005746 immune checkpoint blockade Effects 0.000 description 2
- 238000011532 immunohistochemical staining Methods 0.000 description 2
- 230000002401 inhibitory effect Effects 0.000 description 2
- 210000004698 lymphocyte Anatomy 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 210000002700 urine Anatomy 0.000 description 2
- 206010006002 Bone pain Diseases 0.000 description 1
- 208000009458 Carcinoma in Situ Diseases 0.000 description 1
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 1
- 206010016654 Fibrosis Diseases 0.000 description 1
- 108091008026 Inhibitory immune checkpoint proteins Proteins 0.000 description 1
- 102000037984 Inhibitory immune checkpoint proteins Human genes 0.000 description 1
- 208000008930 Low Back Pain Diseases 0.000 description 1
- 208000002193 Pain Diseases 0.000 description 1
- 102100040678 Programmed cell death protein 1 Human genes 0.000 description 1
- 101710089372 Programmed cell death protein 1 Proteins 0.000 description 1
- 206010000269 abscess Diseases 0.000 description 1
- 208000009956 adenocarcinoma Diseases 0.000 description 1
- 239000002671 adjuvant Substances 0.000 description 1
- 230000004596 appetite loss Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 210000003719 b-lymphocyte Anatomy 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 238000009534 blood test Methods 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 238000007469 bone scintigraphy Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 210000000481 breast Anatomy 0.000 description 1
- 239000012830 cancer therapeutic Substances 0.000 description 1
- 210000003679 cervix uteri Anatomy 0.000 description 1
- 238000002512 chemotherapy Methods 0.000 description 1
- 210000000038 chest Anatomy 0.000 description 1
- 210000001072 colon Anatomy 0.000 description 1
- 238000002574 cystoscopy Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 210000004696 endometrium Anatomy 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000004761 fibrosis Effects 0.000 description 1
- 108020001507 fusion proteins Proteins 0.000 description 1
- 102000037865 fusion proteins Human genes 0.000 description 1
- 210000000232 gallbladder Anatomy 0.000 description 1
- 201000005787 hematologic cancer Diseases 0.000 description 1
- 230000002489 hematologic effect Effects 0.000 description 1
- 208000024200 hematopoietic and lymphoid system neoplasm Diseases 0.000 description 1
- 238000007490 hematoxylin and eosin (H&E) staining Methods 0.000 description 1
- 208000006750 hematuria Diseases 0.000 description 1
- 238000001794 hormone therapy Methods 0.000 description 1
- 210000000987 immune system Anatomy 0.000 description 1
- 238000011293 immunotherapeutic strategy Methods 0.000 description 1
- 201000004933 in situ carcinoma Diseases 0.000 description 1
- 210000003734 kidney Anatomy 0.000 description 1
- 210000000265 leukocyte Anatomy 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 210000004185 liver Anatomy 0.000 description 1
- 235000021266 loss of appetite Nutrition 0.000 description 1
- 208000019017 loss of appetite Diseases 0.000 description 1
- 210000004072 lung Anatomy 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000001394 metastastic effect Effects 0.000 description 1
- 206010061289 metastatic neoplasm Diseases 0.000 description 1
- 230000027939 micturition Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000003205 muscle Anatomy 0.000 description 1
- 230000000771 oncological effect Effects 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 210000001672 ovary Anatomy 0.000 description 1
- 210000000496 pancreas Anatomy 0.000 description 1
- 201000010198 papillary carcinoma Diseases 0.000 description 1
- 238000002600 positron emission tomography Methods 0.000 description 1
- 210000002307 prostate Anatomy 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 210000000664 rectum Anatomy 0.000 description 1
- 210000003491 skin Anatomy 0.000 description 1
- 230000000391 smoking effect Effects 0.000 description 1
- 210000000952 spleen Anatomy 0.000 description 1
- 206010041823 squamous cell carcinoma Diseases 0.000 description 1
- 210000002784 stomach Anatomy 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 230000008961 swelling Effects 0.000 description 1
- 210000001550 testis Anatomy 0.000 description 1
- 210000001685 thyroid gland Anatomy 0.000 description 1
- 238000003325 tomography Methods 0.000 description 1
- 231100000027 toxicology Toxicity 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 206010044412 transitional cell carcinoma Diseases 0.000 description 1
- 210000003932 urinary bladder Anatomy 0.000 description 1
- 208000023747 urothelial carcinoma Diseases 0.000 description 1
- 210000004291 uterus Anatomy 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
- 208000016261 weight loss Diseases 0.000 description 1
- 230000004580 weight loss Effects 0.000 description 1
- 238000007482 whole exome sequencing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
- G16B40/20—Supervised data analysis
Definitions
- Cancer immunotherapy is an emerging discipline of cancer treatments whose main goal is to harness the patient’s own immune system to recognize and destroy tumor cells.
- Various forms of immunotherapy are being developed and are in variable stages of preclinical and clinical development.
- Forms of immunotherapy include, but are not limited to, monoclonal antibodies and adoptive cell transfer.
- Checkpoint inhibition is one aspect of cancer immunotherapy and shows efficacy in both clinical and preclinical studies as an adjuvant and alternative to traditional cancer therapies.
- the effectiveness of checkpoint inhibition results from releasing T cells from the inhibitory effects of checkpoint molecules.
- T cells in the tumor microenvironment (TME) in response to multiple TME-derived factors, increase expression of checkpoint molecules, such as programmed cell death 1 (PD-1) and programmed cell death 1 ligand 1 (PD-L1).
- PD-1 programmed cell death 1
- PD-L1 programmed cell death 1 ligand 1
- Cancer therapeutics including checkpoint inhibitors provide challenges when the therapeutics lack efficacy for a patient, either with primary resistance or acquired resistance.
- Primary resistance can occur when a cancer does not respond to an immunotherapeutic strategy, and in the case of checkpoint inhibitors can occur when the cancer cells fail to express them, thereby providing no target for the therapeutic.
- Embodiments of the disclosure include a system for predicting patient response to an immuno-oncology therapeutic, the system comprising: a data store for storing an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample; a computing device communicatively connected to the data store, the computing device comprising: a first neural network model configured to generate one or more tumor mutation burden (TMB) indicators from the one or more features; a second neural network model configured to generate one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features and/or configured to generate one or more programmed death-1 (PD-1) expression level indicators from the one or more features; a third neural network model configured to calculate one or more tumor infiltrating lymphocyte (TIL)-related metrics from the one or more features; or a fourth neural network model configured to predict a response status of the patient to the immuno-oncology therapeutic in accordance with the one or more TMB indicators, the one or more PD-L1 expression level indicators
- the prediction of the response status of the patient to the immuno-oncology therapeutic is performed in accordance with one or more medical history features of the patient.
- the third neural network model may comprise a segmentation algorithm configured to generate an identification of a tumor region in the tissue sample based on the one or more features, wherein the identification is used to further generate the one or more TIL- related metrics.
- the segmentation algorithm is configured to predict a TIL mask, which is used in combination with a tumor mask to calculate one or more TIL related metrics.
- the one or more TIL-related metrics comprises a predicted number of TIL clusters in the identified tumor region; a predicted size of TIL clusters in the identified tumor region; a predicted TIL cluster spread in the identified tumor region; a total count of TILs in the identified tumor region; distance between TIL cells vs. distance between tumor cells; distance of TIL cells from tumor cells; average size of TIL cells; a stromal TIL% (sTIL%) (which may be calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL); and/or an intra-tumor TIL% (iTTL%), which may be calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL.
- sTIL% stromal TIL%
- iTTL% intra-tumor TIL%
- the first neural network model comprises a classification algorithm and the one or more TMB indicators comprises a TMB indicator for the tissue sample that is either a TMB positive classification or TMB negative classification.
- the classification algorithm may generate the TMB positive classification if the tissue sample has a TMB value that exceeds a pre-set TMB threshold value.
- the first neural network model comprises a regression algorithm and the one or more TMB indicators comprises a TMB indicator that is a quantitative TMB score.
- the second neural network model comprises a classification algorithm and the one or more PD-L1 expression level indicators comprises a PD-L1 expression level indicator that is either a PD- L1 expression positive classification or a PD-L1 expression level negative classification.
- the second neural network model may comprise a classification algorithm and the one or more PD- 1 expression level indicators comprises a PD-1 expression level indicator that is either a PD-1 expression positive classification or a PD-1 expression level negative classification.
- the classification algorithm generates the PD-L1 expression positive classification if the tissue sample has a PD-L1 expression level value that exceeds a pre-set PD-L1 expression level value.
- the classification algorithm generates the PD-1 expression positive classification if the tissue sample has a PD-1 expression level value that exceeds a pre-set PD-1 expression level value.
- the second neural network model may comprise a regression algorithm and the one or more PD-L1 expression level indicators comprises a quantitative PD-L1 score.
- the second neural network model comprises a regression algorithm and the one or more PD-1 expression level indicators comprises a quantitative PD-1 score.
- the quantitative PD-L1 score and/or the quantitative PD- I score may be tumor proportion score (TPS).
- the quantitative PD-L1 score and/or the quantitative PD-1 score may be a combined positive score (CPS).
- the third neural network model is further configured to generate one or more TIL masks from the one or more features; and wherein the fourth neural network model is further configured to predict the response status of the patient to the immunooncology therapeutic using the one or more TIL masks.
- the immuno-oncology therapeutic is a check point inhibitor, and the check point inhibitor may be a PD-L1 inhibitor or PD-1 inhibitor.
- Any tissue sample may be from the patient’s bladder; and in some embodiments the patient response to the immuno-oncology therapeutic is related to bladder cancer.
- the tissue sample may be from a patient known to have PD-L1 -positive cancer.
- the tissue sample may be from a patient known to have PD-1 -positive cancer.
- the tissue sample is from a patient known to have any one or more of the following: non-small cell lung cancer, melanoma, Hodgkin lymphoma, bladder cancer, kidney cancer, and breast cancer.
- the immuno-oncology therapeutic targets PD-L1 or PD-1, in specific embodiments.
- the immuno-oncology therapeutic comprises one or more antibodies, one or more adoptive cell therapies, or a combination thereof.
- the one or more antibodies may comprise at least one of: a monoclonal antibody, a bispecific antibody, or a trispecific antibody.
- the one or more adoptive cell therapies comprises immune cells that express one or more engineered antigen receptors.
- the one or more engineered antigen receptors comprises at least one of: a chimeric antigen receptor, a non-natural T cell receptor, or a combination thereof.
- the one or more adoptive cell therapies comprises at least one of: T cells, natural killer cells, natural killer T cells, or a combination thereof.
- the third neural network model comprises a region/TIL segmentation algorithm and a cell segmentation algorithm.
- the cell segmentation algorithm may calculate one or more cell-based features.
- the one or more cell-based features include morphology, color (nuclear), color (extra-nuclear), texture (energy, correlation, contrast, homogeneity, dissimilarity, or angular second moment).
- the cell segmentation algorithm generates a number of TIL cells or average number of TIL cells.
- the cell segmentation algorithm may generate an average tumor cell size, an average TIL cell size, an average distance between TIL cells, an average color of TIL cells, and/or an average distance between TIL cells vs tumor cells, in certain embodiments.
- Embodiments of the disclosure include methods for predicting patient response to an immuno-oncology therapeutic, the method comprising obtaining, by one or more processors, an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample; generating, by the one or more processors, via a first neural network model, one or more tumor mutation burden (TMB) indicators from the one or more features; generating, by the one or more processors, via a second neural network model, one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features and/or one or more programmed death-1 (PD-1) expression level indicators from the one or more features; generating, by the one or more processors, via a third neural network model, one or more tumor infiltrating lymphocyte (TIL) related metrics from the one or more features; and predicting, by the one or more processors, via a fourth neural network model, a response status of the patient to the immuno-oncology therapeutic using the one or more TMB indicators
- the method further comprises selecting, in accordance with the response status of the patient, whether to include the patient in a cohort of patients participating in a clinical trial. In some embodiments, the method further comprises displaying, on a display screen, a graphical representation indicating the response status of the patient.
- the prediction of the response status of the patient to the immuno-oncology therapeutic further uses one or more medical history features of the patient.
- the third neural network is a segmentation algorithm that identifies a tumor region in the tissue sample, in specific embodiments.
- the segmentation algorithm predicts one or more TIL mask structure features in the identified tumor region to generate the one or more TIL-related metrics.
- the predicted TIL mask structure feature is a predicted number of TIL clusters in the identified tumor region, a predicted size of TIL clusters in the identified tumor region, or a predicted TIL cluster spread in the identified tumor region.
- the one or more TIL mask structure features comprises a total count of TILs in the identified tumor region, distance between TIL cells vs.
- the TIL% value is an intra-tumor TIL% (iTIL%), and the iTIL% is calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL.
- the TIL% value is a stromal TIL% (sTIL%), and the sTIL% value is calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL.
- the first neural network model is a classification algorithm that generates a TMB indicator that is either a TMB positive classification or TMB negative classification for the tissue sample.
- the classification algorithm generates the TMB positive classification if the tissue sample has a TMB value that exceeds a pre-set TMB threshold value, in specific embodiments.
- the first neural network model may be a regression algorithm that generates a TMB indicator that is a quantitative TMB score, in various embodiments
- the second neural network model is a classification algorithm that generates a PD-L1 expression level indicator that is either a PD-L1 expression positive classification or a PD-L1 expression level negative classification, in certain embodiments.
- the second neural network model is a classification algorithm that generates a PD-1 expression level indicator that is either a PD- 1 expression positive classification or a PD- 1 expression level negative classification, in specific embodiments.
- the classification algorithm generates the PD-L1 expression positive classification if the tissue sample has a PD- L1 expression level value that excess a pre-set PD-L1 expression level value.
- the classification algorithm generates the PD-1 expression positive classification if the tissue sample has a PD- I expression level value that excess a pre-set PD-1 expression level value, in specific embodiments.
- the second neural network model is a regression algorithm that generates a PD-L1 expression level indicator that is a quantitative PD-L1 score.
- the second neural network model is a regression algorithm that generates a PD-1 expression level indicator that is a quantitative PD-1 score.
- the quantitative PD-L1 score and/or PD-1 score is tumor proportion score (TPS), in some embodiments.
- the quantitative PD-L1 score and/or PD-1 score is a combined positive score (CPS), in some embodiments.
- the immuno-oncology therapeutic is a check point inhibitor, such as a PD-L1 inhibitor or PD-1 inhibitor, in some embodiments.
- the method further comprises generating, by the one or more processors, via the third neural network model, one or more TIL masks from the one or more features, wherein the predicting of the response status of the patient to the immuno- oncology therapeutic further comprises using the one or more TIL masks.
- the tissue sample is from bladder and the patient response to the immuno-oncology therapeutic is related to bladder cancer. In various embodiments, wherein a determination is made that the patient is or will be a responder based on the response status; and wherein, in accordance with the determination, the patient is administered a therapeutically effective amount of the immuno-oncology therapeutic. In some embodiments, wherein a determination is made that the patient is not or will not be a responder based on the response status; and wherein, in accordance with the determination, the patient is not administered the immuno-oncology therapeutic.
- the tissue sample is from a patient known to have PD-L1 -positive cancer or PD-1 -positive cancer, in specific embodiments.
- the tissue sample is from a patient known to have any of the following: non-small cell lung cancer, melanoma, Hodgkin lymphoma, bladder cancer, kidney cancer, or breast cancer.
- the immuno-oncology therapeutic targets PD-L 1 or PD-
- the immuno-oncology therapeutic comprises one or more antibodies, adoptive cell therapies, immunomodulators, or a combination thereof, in specific embodiments, and the antibody may be a monoclonal antibody, a bispecific antibody, or a trispecific antibody.
- the adoptive cell therapy comprises immune cells that express one or more engineered antigen receptors, in specific embodiments.
- the engineered antigen receptor is a chimeric antigen receptor, a non-natural T cell receptor, or a combination thereof.
- the adoptive cell therapy comprises T cells, natural killer cells, natural killer T cells, or a combination thereof, in various embodiments.
- the method may further comprising obtaining the sample from the patient and/or diagnosing the patient as having cancer.
- a method comprising: obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; generating, by the one or more processors, via one or more neural network models one or more of the following: (a) one or more tumor mutation burden (TMB) indicators from the one or more features; (b) one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death- 1 (PD-1) expression level indicators from the one or more features; and (c) one or more tumor infiltrating lymphocyte (TIL) percentage (%) values from the one or more features; predicting, by the one or more processors, via the one or more neural network models, a response status of the patient to an immuno-oncology therapeutic using one or more of: the one or more TMB indicators,
- a method comprising: obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; and generating, by the one or more processors, via one or more neural network models a prediction of a response status of the patient to an immuno-oncology therapeutic using the image.
- a system for predicting patient response to an immuno-oncology therapeutic comprising: a data store for storing an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample; a computing device communicatively connected to the data store, the computing device comprising: a first neural network model configured to generate one or more tumor mutation burden (TMB) indicators from the one or more features; a second neural network model configured to generate one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features and/or configured to generate one or more programmed death- 1 (PD-1) expression level indicators from the one or more features; a third neural network model configured to calculate one or more tumor infiltrating lymphocyte (TIL)- related metrics from the one or more features; a fourth neural network model configured to calculate one or more cell-based features from the one or more features; and a fifth neural network model configured to predict a response status of the patient to the immuno-oncology therapeutic in
- there is a method comprising: obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; and training, by the one or more processors, one or more neural network models to predict a response status of the patient to an immuno-oncology therapeutic using the image.
- a method for predicting patient response to an immuno-oncology therapeutic comprising: obtaining, by one or more processors, an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample; generating, by the one or more processors, one or more tumor mutation burden (TMB) indicators from the one or more features (optionally); generating, by the one or more processors, from the one or more features, one or more programmed death-ligand 1 (PD-L1) expression level indicators or one or more programmed death- 1 (PD-1) expression level indicators; generating, by the one or more processors, one or more tumor infiltrating lymphocyte (TIL) related metrics from the one or more features; and training, by the one or more processors, a fourth neural network model to predict a response status of the patient to the immuno-oncology therapeutic based on one or more of: the one or more TMB indicators (when utilized), the one or more PD-L1 expression level indicators, the one
- Figure 1A depicts a system for predicting a patient or treatment response to a therapeutic, in accordance with various embodiments.
- Figure IB depicts another system for predicting a patient or treatment response to a therapeutic, in accordance with various embodiments.
- Figure 1 C depicts another system for predicting a patient or treatment response to a therapeutic, in accordance with various embodiments
- Figure 2A illustrates a method for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- Figure 2B illustrates another method for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- Figure 2C illustrates another method for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- Figure 2D illustrates a method for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- Figure 2E illustrates another method for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- Figure 2F illustrates another method for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- Figure 3 is a block diagram of a computing system configured to perform a method of predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- Embodiments of the disclosure include methods in which it is predicted whether or not an individual will respond to one or more certain therapies.
- the one or more certain therapies comprise one or more immuno-oncology therapeutics of any kind.
- the one or more immuno-oncology therapeutics target one or more checkpoint inhibitors.
- the one or more checkpoint inhibitors comprises PD-L1 or PD-1, or a mixture thereof.
- the type of immuno-oncology therapeutic may be of any kind, including a therapeutic that utilizes, at least in part, an antibody of any kind.
- the antibody comprises a monoclonal antibody and may or may not be humanized.
- the antibody may be utilized as an immuno-oncology therapeutic by itself, or the antibody may be configured as part of a chimeric molecule of any kind, such as an engineered antigen receptor.
- the antibody may be configured as part of a fusion protein of any kind.
- an individual such as a patient, is subject to methods in which multiple parameters are measured in order to determine if the individual will respond to a particular therapy.
- the parameters include one or more tumor mutation burden indicators, one or more PD-L1 expression level indicators and/or one or more PD-1 expression level indicators; and one or more TIL metric values.
- patient characteristics such as biological sex, age, individual or family health history (such as whether or not the individual and or one or more family members has had any disease including cancer before, including PD-L1 -positive cancer and/or PD-1 -positive cancer), and so forth, may also be included as a parameter or parameters to consider in order to determine if the individual will respond to a particular therapy.
- an individual such as a patient, is predicted to be or determined to be a responder to one or more immuno-oncology therapeutics.
- an individual is predicted to be, or determined to be, at a higher probability of responding to one or more immuno-oncology therapeutics when compared to a particular population, such as the general population, a population of cancer patients, a population that lacks expression of one or more checkpoints, such as PD-L1 or PD-1, and so forth.
- Figure 1A depicts a system 100 for predicting a patient or treatment response to a therapeutic, in accordance with various embodiments.
- the system 100 may be applicable to any suitable therapeutics, such as for example, for cancers of any kind, the system 100 depicted in Figure 1A is illustrated for predicting a patient response to an immunooncology therapeutic.
- the system 100 includes one or more neural networks 110 that can be used in analyzing an image 105 to make predictions of a patient response 190 (also referred to herein as “response 190”) based on the analysis of the image 105.
- one or more patient features 108 may also be used in the analysis and prediction of the patient response 190 to the therapeutic.
- the image 105 can be an image of a tissue sample from the patient.
- the image 105 can include one or more features of the tissue sample (e.g., shapes, sizes, distribution, and/or density of cells, etc.), which can be used in the analysis by the one or more neural networks 110 to predict the patient response 190 to the therapeutic.
- the image 105 can include a Hematoxylin and Eosin (H&E)-stained tissue sample or any other samples from a patient that can be used in routine histopathological techniques for diagnosing a disease, such as cancer, including bladder cancer.
- H&E Hematoxylin and Eosin
- the image 105 can be acquired from a lab or a clinical department, such as a histology department, prior to analysis, or can be stored in a data store, database, or data storage for retrieving prior to analysis.
- the image 105 may be of tissue that is fresh or that was frozen.
- the one or more patient features 108 can include any feature related and/or pertinent to the patient that can help with the analysis or the prediction of the patient response to a therapeutic.
- Such features 108 may include, for example, but not limited to, age, biological sex, weight, smoking history, one or more symptoms, previous lines of treatment, previous treatment plans, time since diagnosis, the number of previous diagnoses to the same disease or any other diseases that may affect related organs or tissues, patient family medical information, presence or absence of one or more biomarkers for the disease, or any anatomical portions of the patient.
- the one or more patient features 108 may include one or more medical history features of the patient.
- the one or more patient features 108 may be extracted from a patient health profde, for example, from an electronic health record system.
- the system 100 includes one or more neural networks 110 for analysis of input images, such as image 105 and for predicting the patient response 190 based on the analysis.
- the one or more neural networks 110 include a first neural network model 120 (also referred to herein as “model 120”), a second neural network model 140 (also referred to herein as “model 140”), and a third neural network model 160 (also referred to herein as “model 160”).
- Each of the models 120, 140, and/or 160 can include a machine learning algorithm that can input the image 105, along with the one or more features of the tissue sample, to generate respective outputs 130, 150, and/or 170, as illustrated in Figure 1A.
- the one or more neural networks 110 includes a fourth neural network model 180 (also referred to herein as “model 180”), which can include a machine learning algorithm that can ingest one or more of the outputs 130, 150, and 170 as input into the model 180 to generate the patient response 190.
- model 180 can input the image 105 and/or the one or more features of the tissue sample, as illustrated in Figure 1 A.
- the model 180 can also input the one or more patient features 108 as part of the analysis to generate the patient response 190, as illustrated in Figure 1A.
- the one or more neural networks 110 may include, for example, without limitation, at least one, or any combination, of a Convolutional Neural Network (CNN), a Feedforward Neural Network (FNN), a Recurrent Neural Network (RNN), a Modular Neural Network (MNN), a Residual Neural Network (ResNet), an Ordinary Differential Equations Neural Networks (neural-ODE), or another type of neural network.
- CNN Convolutional Neural Network
- FNN Feedforward Neural Network
- RNN Recurrent Neural Network
- MNN Modular Neural Network
- Residual Neural Network Residual Neural Network
- Neural-ODE Ordinary Differential Equations Neural Networks
- Each of the one or more neural networks 110, and thus, by extension, the models 120, 140, 160, and 180 can include one or more algorithms, one or more mathematical techniques, one or more machine learning algorithms, or a combination thereof.
- the one or more neural networks 110 may be trained using large amounts of datasets.
- the training of the one or more neural networks 110, and thus, by extension, the models 120, 140, 160, and 180 can be self-supervised learning (SSL) or in supervised fashion.
- the models 120, 140, 160, and 180 may be pretrained using SSL and subsequently trained for one or more specific tasks (e.g., various indicators or metrics listed below) via supervised learning.
- the datasets for training the models can include annotated images by clinicians/pathologists.
- each of the trained models 120, 140, 160, and/or 180 can then be configured to parse the one or more features of the tissue sample from the image 105, for example, to extract necessary data, learn from it, and then respectively make a determination or prediction, such as the outputs 130, 150, 170, and/or the patient response 190.
- the model 120 can be configured to generate one or more tumor mutation burden (TMB) indicators as the output 130, based on the one or more features of the image 105, in accordance with various embodiments.
- the model 140 can be configured to generate one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or configured to generate one or more programmed death-1 (PD-1) expression level indicators as the output 150 from the one or more features of the image 105, in accordance with various embodiments.
- the immuno-oncology therapeutic is a checkpoint inhibitor, wherein the checkpoint inhibitor can be a PD-L1 inhibitor or PD-1 inhibitor.
- the model 160 can be configured to calculate one or more tumor infiltrating lymphocyte (TIL)-related metrics as the output 170 from the one or more features of the image 105, in accordance with various embodiments.
- the model 120 includes a classification algorithm or a regression algorithm, both of which are trained using the annotated dataset as described above.
- the classification algorithm of the model 120 can ingest one or more features for the tissue sample from the image 105 to generate the output 130 as either a TMB positive classification or TMB negative classification.
- the TMB positive classification is generated as the output 130 when a TMB value exceeds a pre-set TMB threshold value.
- the regression algorithm of the model 120 can generate a TMB value or a quantitative TMB score as the TMB indicator for the output 130.
- the model 140 includes a classification algorithm or a regression algorithm, both of which are trained using the annotated dataset as described above.
- a classification algorithm or a regression algorithm can be employed to perform the analysis.
- the classification algorithm of the model 140 can generate a PD-L1 expression level indicator as the output 150 that is either a PD-L1 expression positive classification or a PD-L1 expression level negative classification.
- the classification algorithm of the model 140 can generate the PD-L1 expression positive classification if the tissue sample of the image 105 has a PD-L1 expression level value that exceeds a pre-set PD-L1 expression level value.
- the regression algorithm of the model 140 can generate one or more PD-L1 expression level indicators as the output 150 that includes a quantitative PD-L1 score.
- the quantitative PD-L1 score can be tumor proportion score (TPS) or a combined positive score (CPS).
- the classification algorithm of the model 140 can generate a PD-1 expression level indicator as the output 150 that is either a PD-1 expression positive classification or a PD-1 expression level negative classification.
- the classification algorithm of the model 140 can generate the PD-1 expression positive classification if the tissue sample of the image 105 has a PD-1 expression level value that exceeds a pre-set PD-1 expression level value.
- the regression algorithm of the model 140 can generate the one or more PD-1 expression level indicators as the output 150 that includes a quantitative PD-1 score.
- the quantitative PD-1 score can be tumor proportion score (TPS) or a combined positive score (CPS).
- the model 160 may include one or more segmentation algorithms, each of which can be configured to generate one or more TIL masks, one or more tumor masks, one or more stroma masks, one or more necrosis masks, or one or more cell nuclei masks as the output 170 from the one or more features of the image 105.
- the model 160 may include a region segmentation algorithm.
- the model 160 may include a cell segmentation algorithm.
- the model 160 can calculate TIL-related metrics, such as for example, but not limited to, the TIL%, sTIL%, or the number of TIL clusters, etc. , by using the one or more TIL masks, one or more tumor masks, one or more storm masks, one or more necrosis masks, or one or more cell nuclei masks as the output 170.
- the model 160 may include a segmentation algorithm, which can be trained using the annotated dataset as described above.
- the segmentation algorithm may be employed to generate an identification of a tumor region in the tissue sample based on the one or more features of the image 105, in accordance with various embodiments. This identification may be used to further generate or calculate one or more TIL-related metrics as the output 170 from the one or more features of the image 105.
- the segmentation algorithm may be referred to as a region segmentation algorithm or a TIL region segmentation algorithm.
- the segmentation algorithm may be referred to as a tumor region segmentation algorithm.
- the segmentation algorithm may be configured to predict a TIL mask, which is used in combination with a tumor mask to calculate one or more TIL-related metrics as the output 170.
- the one or more TIL-related metrics may include a predicted number of TIL clusters in the identified tumor region as the output 170.
- the one or more TIL-related metrics may include a predicted size of TIL clusters in the identified tumor region as the output 170.
- the one or more TIL- related metrics may include a predicted TIL cluster spread in the identified tumor region as the output 170.
- the one or more TIL-related metrics may include an intratumor TIL% (iTIL%) as the output 170.
- the iTIL% may be calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL as the output 170.
- the one or more TIL-related metrics may include a stromal TIL% (sTIL%) as the output 170.
- the sTIL% value may be calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL as the output 170.
- the model 160 may include a segmentation algorithm for cell segmentation, which can be referred to as a cell segmentation algorithm.
- the cell segmentation algorithm may input the image 105 and perform segmentation of the cells as the output 170.
- the segmented cells may be labeled based on the region segmentation output.
- the segmented cell is labeled as a TIL cell if the cell is in a TIL region.
- the cell segmentation algorithm is configured to calculate one or more features related to the cells, also referred to herein as cell-based features, contained in the image 105 of the tissue sample.
- the calculated cell-based features may include morphology, color (nuclear), color (extra-nuclear), texture (energy, correlation, contrast, homogeneity, dissimilarity, angular second moment), chromatin skeleton morphology and color, among many other features that can be extracted.
- the cell segmentation algorithm of the model 160 may be configured to generate a number of TIL cells or average number of TIL cells as the output 170. In various embodiments, the cell segmentation algorithm of the model 160 may be configured to generate an average tumor cell size as the output 170. In various embodiments, the cell segmentation algorithm of the model 160 may be configured to generate an average TIL cell size as the output 170. In various embodiments, the cell segmentation algorithm of the model 160 may be configured to generate an average distance between TIL cells as the output 170. In various embodiments, the cell segmentation algorithm of the model 160 may be configured to generate an average color of TIL cells as the output 170.
- the cell segmentation algorithm of the model 160 may be configured to generate an average distance between TIL cells vs tumor cells, etc, as the output 170.
- the one or more TIL-related metrics may include a total count of TILs in the identified tumor region as the output 170.
- the one ormore TIL-related metrics may include distance between TIL cells vs. distance between tumor cells as the output 170.
- the one or more TIL-related metrics may include a distance of TIL cells from tumor cells as the output 170.
- the one or more TIL- related metrics may include average color of TIL cells as the output 170.
- the one or more TIL-related metrics may include average size of TIL cells as the output 170.
- the model 180 can be configured to generate a response status, for example, of the patient or treatment response 190 to the immuno-oncology therapeutic by applying one or more of the outputs 130, 150, and 170 as input into the model 180.
- the one or more TMB indicators of the output 130, the one or more PD-L1 expression level indicators or one or more PD-1 expression level indicators, respectively, of the output 150, and the one or more TIL-related metrics of the output 170 can be used in the analysis by the model 180 to predict a response status of the patient to the immuno-oncology therapeutic, in accordance with various embodiments.
- the model 180 can include a machine learning algorithm that is trained using the patient’s immune-oncology treatment response data, wherein the response 190 can include annotated labels of responder, partial responder, or non-responder as the ground truth for training.
- the model 180 can be configured such that each of the outputs 130, 150, and 170 are weighed against one another in the analysis by the machine learning algorithm of the model 180.
- the different weights applied to each of the outputs 130, 150, and 170 can affect and/or significantly alter the output 190 of the patient response to the therapeutic, such as the immune-oncology therapeutic.
- the model 180 can be trained using a set of H&E images from patients with response data available for the desired treatment or therapeutic, such as the immune -oncology therapeutic.
- the model 180 can be trained using the following four patients.
- each of the models 120, 140, and 160 can be run to generate the following datasets from the image 105, the outputs 130 (TMB score), 150 (PD-L1 score), and 170 (TIL mask and/or TIL features), respectively:
- the model 180 can be trained as follows: “H&E image, TMB score, PD-L1 score, TIL masks, TIL features,” along with the ground truth or label as either “responder” or “non-responder”, or in some instances, as “partial responder”. Once the model 180 is trained, the model 180 can be used to generate the patient response 190 based on a single image similar to the image 105 and the outputs that are generated by the trained models 120, 140, and 160 using one or more features of the single image.
- Figure IB depicts a system 102 for predicting a patient or treatment response to a therapeutic, in accordance with various embodiments.
- the system 102 is similar to the system 100 described with respect to Figure 1A where the image 105, the neural networks 110 having the models 120 and 140 generating the outputs 130 and 150, respectively, and the model 180 having to ingest the outputs 130 and 150, and/or the image 105, and/or the patient features 108 in order to generate the response 190.
- the system 102 is different to the system 100 in that instead of having the model 160 generate the output 170 that can be ingested by the model 180 to generate the response 190 (as in the system 100), the system 102 includes two models 160a and 160b, which respectively generate outputs 170a and 170b, which can be ingested by the model 180 to generate the response 190.
- the models 160a and 160b represent two separate segmentation algorithms.
- the model 160a may include a region/TIL segmentation algorithm.
- the model 160b may include a cell segmentation algorithm.
- the region/TIL segmentation algorithm of the model 160a may be employed to generate an identification of a tumor region in the tissue sample based on the one or more features of the image 105. In various embodiments, the identification may be used to further generate or calculate one or more TIL-related metrics as the output 170a from the one or more features of the image 105. In various embodiments, the region/TIL segmentation algorithm of the model 160a may be configured to predict a TIL mask, which is used in combination with a tumor mask to calculate one or more TIL-related metrics as the output 170a. In various embodiments, the one or more TIL-related metrics may include a predicted number of TIL clusters in the identified tumor region as the output 170a.
- the one or more TIL-related metrics may include a predicted size of TIL clusters in the identified tumor region as the output 170a. In various embodiments, the one or more TIL-related metrics may include a predicted TIL cluster spread in the identified tumor region as the output 170a. In various embodiments, the one or more TIL-related metrics may include an intra-tumor TIL% (iTIL%) as the output 170a. In various embodiments, the iTTL% may be calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL as the output 170a. In various embodiments, the one or more TIL-related metrics may include a stromal TIL% (sTIL%) as the output 170a. In various embodimentsTIL% value may be calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL as the output 170a.
- the cell segmentation algorithm of the model 160b may input the image 105 and perform segmentation of the cells as the output 170b.
- the cell segmentation algorithm of the model 160b may input a region/TIL mask generated by the model 160a and perform segmentation of the cells as the output 170b.
- the segmented cells may be labeled based on the region segmentation output. For example, the segmented cell is labeled as a TIL cell if the cell is in a TIL region.
- the cell segmentation algorithm is configured to calculate one or more features related to the cells, also referred to herein as cell -based features, contained in the image 105 of the tissue sample.
- the calculated cell -based features may include morphology, color (nuclear), color (extra-nuclear), texture (energy, correlation, contrast, homogeneity, dissimilarity, angular second moment), chromatin skeleton morphology and color, among many other features that can be extracted.
- the cell segmentation algorithm of the model 160b may be configured to generate a number of TIL cells or average number of TIL cells as the output 170b. In various embodiments, the cell segmentation algorithm of the model 160b may be configured to generate an average tumor cell size as the output 170b. In various embodiments, the cell segmentation algorithm of the model 160b may be configured to generate an average TIL cell size as the output 170b.
- the cell segmentation algorithm of the model 160b may be configured to generate an average distance between TIL cells as the output 170b. In various embodiments, the cell segmentation algorithm of the model 160b may be configured to generate an average color of TIL cells as the output 170b. In various embodiments, the cell segmentation algorithm of the model 160b may be configured to generate an average distance between TIL cells vs tumor cells, etc., as the output 170b. In various embodiments, the one or more TIL-related metrics may include a total count of TILs in the identified tumor region as the output 170b. In various embodiments, the one or more TIL-related metrics may include distance between TIL cells vs. distance between tumor cells as the output 170b.
- the one or more TIL-related metrics may include a distance of TIL cells from tumor cells as the output 170b. In various embodiments, the one or more TIL-related metrics may include average color of TIL cells as the output 170b. In various embodiments, the one or more TIL-related metrics may include average size of TIL cells as the output 170b.
- the model 180 can be configured to generate a response status, for example, of the patient or treatment response 190 to the immuno-oncology therapeutic by applying one or more of the outputs 130, 150, 170a and 170b as input into the model 180.
- the one or more TMB indicators of the output 130, the one or more PD-L1 expression level indicators or one or more PD-1 expression level indicators, respectively, of the output 150, and the one or more TIL-related metrics of the output 170 can be used in the analysis by the model 180 to predict a response status of the patient to the immuno-oncology therapeutic, in accordance with various embodiments.
- the model 180 can include a machine learning algorithm that is trained using the patient’s immune-oncology treatment response data, wherein the response 190 can include annotated labels of responder, partial responder, or non-responder as the ground truth for training.
- the model 180 can be configured such that each of the outputs 130, 150, 170a, and 170b are weighed against one another in the analysis by the machine learning algorithm of the model 180.
- the different weights applied to each of the outputs 130, 150, 170a, and 170b can affect and/or significantly alter the output 190 of the patient response to the therapeutic, such as the immune-oncology therapeutic.
- each of the models 120, 140, 160a, and 160b can be run to generate the following datasets from the image 105, the outputs 130 (TMB score), 150 (PD-L1 score), 170a (TIL mask and/or TIL features), and 170b (cell-based features), respectively:
- the model 180 can be trained as follows: “H&E image, TMB score, PD-L1 score, TIL masks, TIL features, cell -based features” along with the ground truth or label as either “responder” or “non-responder”, or in some instances, as “partial responder”.
- the model 180 can be used to generate the patient response 190 based on a single image similar to the image 105 and the outputs that are generated by the trained models 120, 140, 160a, and 160b using one or more features of the single image.
- Figure 1 C depicts another system 200 for predicting a patient or treatment response to a therapeutic, in accordance with various embodiments. Similar to the system 100 as described with respect to Figure 1A, the system 200 may be applicable to any suitable therapeutics, such as for example, for any type of cancer. Although the system 200 is depicted in Figure 1C for predicting a patient response to an immuno-oncology therapeutic, the system 200 is suitable for generating a patient response to any other therapeutics. As shown in Figure 1C, the system 200 includes one or more neural networks 210 that can be used in analyzing an image 205 to make predictions of a patient response 290 (also referred to herein as “response 290”) based on the analysis of the image 205.
- a patient response 290 also referred to herein as “response 290”
- one or more patient features 208 may also be used in the analysis and prediction of the patient response 290 to the therapeutic. Since the image 205 and the one or more patient features 208 may be substantially similar or identical to the image 105 and the one or more patient features 108, respectively, described with respect to Figure 1A, they will not be described in further detail.
- the system 200 includes one or more neural networks 210 for analysis of input images, such as image 205 and for predicting the patient response 290 based on the analysis.
- the one or more neural networks 210 include one or more classification algorithms, one or more regression algorithms, and/or one or more segmentation algorithms for analysis and generating one or more final outputs and intermediatory outputs.
- the intermediatory outputs may be generated by one or more of the classification algorithms, regression algorithms, and/or segmentation algorithms, and then used as inputs in another one of the one or more of the classification algorithms, regression algorithms, and/or segmentation algorithms to generate the final output(s).
- Each of the classification algorithms, regression algorithms, and/or segmentation algorithms of the one or more neural networks 210 is a machine learning algorithm that can input the image 205, along with the one or more features of the tissue sample, to generate the response 290, as illustrated in Figure 1C.
- the one or more neural networks 210 may include, for example, without limitation, at least one of a Convolutional Neural Network (CNN), a Feedforward Neural Network (FNN), a Recurrent Neural Network (RNN), a Modular Neural Network (MNN), a Residual Neural Network (ResNet), an Ordinary Differential Equations Neural Networks (neural-ODE), or another type of neural network, or any combination thereof.
- the one or more neural networks 210 may be trained using large amounts of datasets.
- the training of the one or more neural networks 210 can be self-supervised learning (SSL) or in supervised fashion.
- the datasets for training the models can include annotated images by clinicians/pathologists.
- each of the trained classification algorithms, regression algorithms, and/or segmentation algorithms of the one or more neural networks 210 can then be configured to parse the one or more features of the tissue sample from the image 105, for example, to extract necessary data, learn from it, and then respectively make a determination or prediction, such as the patient response 190.
- one or more neural network models 110 or 210 can be configured to generate a prediction of a response status of the patient to an immuno-oncology therapeutic using the image 105 or image 205.
- the one or more neural networks 210 can be configured to generate one or more tumor mutation burden (TMB) indicators based on the one or more features of the image 205, in accordance with various embodiments.
- TMB tumor mutation burden
- the one or more neural networks 210 can be configured to generate one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or configured to generate one or more programmed death- 1 (PD-1) expression level indicators from the one or more features of the image 205, in accordance with various embodiments.
- the one or more neural networks 210 can be configured to calculate one or more tumor infiltrating lymphocyte (TIL) related metrics from the one or more features of the image 205, in accordance with various embodiments.
- TIL tumor infiltrating lymphocyte
- the one or more neural networks 210 are trained using the annotated dataset as described above.
- a classification algorithm of the one or more neural networks 210 can ingest one or more features for the tissue sample from the image 205 to generate a TMB positive classification or TMB negative classification. These features may be used order to generate the one or more TMB indicators.
- the TMB positive classification is generated as an intermediatory output when a TMB value exceeds a pre-set TMB threshold value.
- the regression algorithm of the one or more neural networks 210 can generate a TMB value or a quantitative TMB score as the TMB indicator for the output 130.
- the one or more neural networks 210 can generate one or more PD-L1 expression level indicators or one or more PD-1 expression level indicators.
- the one or more neural networks 210 can include a classification algorithm that can generate a PD-L1 expression level indicator as the output 150 that is either a PD-L1 expression positive classification or a PD-L1 expression level negative classification.
- the classification algorithm can generate the PD-L1 expression positive classification if the tissue sample of the image 105 has a PD-L1 expression level value that exceeds a pre-set PD-L1 expression level value.
- the one or more neural networks 210 can include a regression algorithm to generate one or more PD-L1 expression level indicators as the output 150 that includes a quantitative PD-L1 score.
- the quantitative PD-L1 score can be tumor proportion score (TPS) or a combined positive score (CPS).
- the classification algorithm can generate a PD-1 expression level indicator as an intermediatory output that is either a PD-1 expression positive classification or a PD-1 expression level negative classification.
- the classification algorithm can generate the PD-1 expression positive classification if the tissue sample of the image 205 has a PD-1 expression level value that exceeds a pre-set PD-1 expression level value.
- the regression algorithm can generate the one or more PD-1 expression level indicators that include a quantitative PD-1 score.
- the quantitative PD-1 score can be tumor proportion score (TPS) or a combined positive score (CPS).
- the one or more neural networks 210 can include a segmentation algorithm, which is trained using the annotated dataset as described above.
- the segmentation algorithm is employed to generate an identification of a tumor region in the tissue sample based on the one or more features of the image 205, in accordance with various embodiments. These metrics may be used to generate or calculate one or more TIL-related metrics from the one or more features of the image 205.
- the segmentation algorithm uses one or more TIL mask structure features in the identified tumor region to generate one or more TIL-related metrics as an intermediatory output.
- the one or more TIL mask structure features may include a predicted number of TIL clusters in the identified tumor region as an intermediatory output.
- the one or more TIL mask structure features may include a predicted size of TIL clusters in the identified tumor region as an intermediatory output. In various embodiments, the one or more TIL mask structure features may include a predicted TIL cluster spread in the identified tumor region as an intermediatory output. In various embodiments, the one or more TIL mask structure features may include a total count of TILs in the identified tumor region as an intermediatory output. In various embodiments, the one or more TIL mask structure features may include distance between TIL cells vs. distance between tumor cells as an intermediatory output. In various embodiments, the one or more TIL mask structure features may include a distance of TIL cells from tumor cells as an intermediatory output.
- the one or more TIL mask structure features may include average color of TIL cells as an intermediatory output. In various embodiments, the one or more TIL mask structure features may include average size of TIL cells as an intermediatory output.
- the one or more TIL-related metrics may include an intra-tumor TIL% (iTIL%) as an intermediatory output. In various embodiments, the iTIL% may be calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL an intermediatory output. In various embodiments, the one or more TIL-related metrics may include a stromal TIL% (sTIL%) as an intermediatory output. In various embodiments, the sTTL% value may be calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL as an intermediatory output.
- the one or more neural networks 210 may include one or more segmentation algorithms, each of which can be configured to generate one or more TIL masks, one or more tumor masks, one or more storm masks, one or more necrosis masks, or one or more cell nuclei masks as intermediatory outputs.
- the one or more segmentation algorithms of the one or more neural networks 210 can calculate TIL- related metrics, such as for example, but not limited to, the TIL%, sTIL%, or the number of TIL clusters, etc., by using the one or more TIL masks, one or more tumor masks, one or more storm masks, one or more necrosis masks, or one or more cell nuclei masks.
- the one or more neural networks 210 can be configured to generate a response status, for example, of the patient or treatment response 290 to the immuno-oncology therapeutic by applying one or more of the intermediatory outputs described above as input into the one or more neural networks 210.
- the one or more TMB indicators, the one or more PD-L1 expression level indicators or one or more PD- 1 expression level indicators, and the one or more TIL-related metrics can be used in the analysis to predict a response status of the patient to the immuno-oncology therapeutic, in accordance with various embodiments.
- the one or more neural networks 210 can include a machine learning algorithm that is trained using the patient’s immune- oncology treatment response data, wherein the response 290 can include annotated labels of responder, partial responder, or non-responder as the ground truth for training.
- the one or more neural networks 210 can be configured such that each of the intermediatory outputs are weighed against one another in the analysis by the machine learning algorithms.
- the different weights applied to various intermediatory outputs can affect and/or significantly alter the patient response 290 to the therapeutic, such as the immune-oncology therapeutic.
- no biomarkers are generated.
- the one or more neural networks 210 directly predicts a patient’s response from the image 205.
- FIG. 2A illustrates a method SI 00 for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- the method S 100 can be implemented for example, via a computing system 300 as described with respect to Figure 3 below.
- the method SI 00 includes at step SI 10, obtaining or receiving, by one or more processors of the computing system 300, an image of a tissue sample, such as the image 105, from a patient; at step S120, generating, by the one or more processors, via a first neural network model, such as the model 120, one or more tumor mutation burden (TMB) indicators (e.g., as the output 130) from the one or more features; at step S130, generating, by the one or more processors, via a second neural network model, such as the model 140, one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death-1 (PD-1) expression level indicators (e.g., as the output 150)
- TMB tumor
- the method SI 00 can optionally include displaying, on a display screen, a graphical representation indicating the response status of the patient. For instance, predicted response statuses of individuals may be displayed to a clinician or other health care provider, a clinical trial administrator, a researcher, or other entity via a graphical user interface, for example as part of a clinical trial management software or other software.
- the method SI 00 can optionally include selecting, in accordance with the response status of the patient, whether to include the patient in a cohort of patients participating in a clinical trial. In specific embodiments, it may be desirable to determine whether or not a candidate for a clinical trial would be able to respond to the immuno-oncology therapeutic.
- Certain embodiments include treatment of an individual with a therapeutically effective amount of an immuno-oncology therapeutic wherein the individual is a responder as determined by systems or methods disclosed herein, and the individual may or may not be a participant in a clinical trial.
- an individual that is predicted to be a responder based on measurements, assaying, and/or generation of multiple neural network models as encompassed herein is administered a therapeutically effective amount of an immuno-oncology therapeutic.
- the prediction of the response status of the patient to the immuno-oncology therapeutic further uses one or more medical history features of the patient.
- the third neural network is a segmentation algorithm that identifies a tumor region in the tissue sample.
- the segmentation algorithm predicts a TIL mask, which is used in combination with a tumor mask to calculate the one or more TIL-related metrics.
- the one or more TIL-related metrics includes a predicted number of TIL clusters in the identified tumor region.
- the one or more TIL-related metrics includes a predicted size of TIL clusters in the identified tumor region.
- the TIL-related metrics includes an intra-tumor TIL% (iTIL%).
- the iTIL% is calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL.
- the TIL% value is a stromal TIL% (sTIL%).
- the sTIL% value is calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL.
- the third neural network is a cell segmentation algorithm calculates one or more cell -based features.
- the one or more cell-based features include morphology, color (nuclear), color (extra-nuclear), texture (energy, correlation, contrast, homogeneity, dissimilarity, angular second moment), or chromatin skeleton morphology and color.
- the cell segmentation algorithm generates a number of TIL cells or average number of TIL cells.
- the cell segmentation algorithm generates an average tumor cell size .
- the cell segmentation algorithm generates an average TIL cell size.
- the cell segmentation algorithm generates an average distance between TIL cells.
- the cell segmentation algorithm generates an average color of TIL cells. In various embodiments, the cell segmentation algorithm generates an average distance between TIL cells vs tumor cells. In various embodiments, the one or more TIL- related metrics includes a predicted TIL cluster spread in the identified tumor region. In various embodiments, the one or more TIL-related metrics comprises a total count of TILs in the identified tumor region. In various embodiments, the one or more TIL-related metrics comprises distance between TIL cells vs. distance between tumor cells. In various embodiments, the one or more TIL-related metrics comprises distance of TIL cells from tumor cells. In various embodiments, the one or more TIL-related metrics comprises average color of TIL cells. In various embodiments, the one or more TIL-related metrics comprises average size of TIL cells.
- the first neural network model is a classification algorithm that generates a TMB indicator that is either a TMB positive classification or TMB negative classification for the tissue sample.
- the classification algorithm generates the TMB positive classification if the tissue sample has a TMB value that exceeds a pre-set TMB threshold value.
- the first neural network model is a regression algorithm that generates a TMB indicator that is a quantitative TMB score.
- the second neural network model is a classification algorithm that generates a PD-L1 expression level indicator that is either a PD- L1 expression positive classification or a PD-L1 expression level negative classification.
- the second neural network model is a classification algorithm that generates a PD-1 expression level indicator that is either a PD-1 expression positive classification or a PD-1 expression level negative classification.
- the classification algorithm generates the PD-L1 expression positive classification if the tissue sample has a PD-L1 expression level value that excess a pre-set PD-L1 expression level value.
- the classification algorithm generates the PD-1 expression positive classification if the tissue sample has a PD-1 expression level value that excess a pre-set PD-1 expression level value.
- the second neural network model is a regression algorithm that generates a PD-L1 expression level indicator that is a quantitative PD-L1 score.
- the second neural network model is a regression algorithm that generates a PD-1 expression level indicator that is a quantitative PD-1 score.
- the quantitative PD-L1 score is a tumor proportion score (TPS) or a combined positive score (CPS).
- the quantitative PD-1 score is a tumor proportion score (TPS) or a combined positive score (CPS).
- the immuno-oncology therapeutic is a check point inhibitor.
- the check point inhibitor is a PD-L1 inhibitor or PD-1 inhibitor.
- the tissue sample is from bladder and the patient response to the immuno-oncology therapeutic is related to bladder cancer.
- a determination is made that the patient is or will be a responder based on the response status.
- the patient is administered a therapeutically effective amount of the immuno-oncology therapeutic.
- the tissue sample is from a patient known to have or suspected of having PD-L1 -positive cancer or PD-1 -positive cancer.
- the tissue sample is from a patient known to have or suspected of having non-small cell lung cancer, melanoma, Hodgkin lymphoma, bladder cancer, kidney cancer, or breast cancer.
- the immuno-oncology therapeutic targets PD-L1 or PD-1.
- the immuno-oncology therapeutic comprises one or more antibodies, adoptive cell therapies, immunomodulators, or a combination thereof.
- the antibody is a monoclonal antibody, a bispecific antibody, or a trispecific antibody.
- the adoptive cell therapy comprises immune cells that express one or more engineered antigen receptors.
- the engineered antigen receptor is a chimeric antigen receptor, a non-natural T cell receptor, or a combination thereof.
- the adoptive cell therapy comprises T cells, natural killer cells, natural killer T cells, or a combination thereof.
- the method SI 00 may optionally include the step of obtaining the sample from the patient. In various embodiments, the method SI 00 may include the step of diagnosing the patient as having cancer.
- FIG. 2B illustrates another method S200 for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- the method S200 can be implemented for example, via a computing system 300 as described with respect to Figure 3 below.
- the method S200 includes at step S210, obtaining or receiving, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; at step S220, generating, by the one or more processors of the computing system 300, via one or more neural network models: (a) one or more tumor mutation burden (TMB) indicators from the one or more features; (b) one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death-1 (PD-1) expression level indicators from the one or more features; and (c) one or more tumor infiltrating lymphocyte (TIL) related metrics from the one or more features; at step S230, predicting, by the one or more TIL
- step S220 can be omitted and step S230 may be directed to a prediction of a response status of the patient to the immune-oncology therapeutic from the image itself without the use of any calculated/generated biomarkers.
- the prediction of the response status of the patient to the immuno-oncology therapeutic further uses one or more medical history features of the patient.
- the one or more neural network models can include a segmentation algorithm configured to predict a TIL mask, which is used in combination with a tumor mask to calculate one or more TIL-related metrics.
- the TIL-related metrics is a predicted number of TIL clusters in the identified tumor region.
- the TIL-related metrics is a predicted size of TIL clusters in the identified tumor region.
- the TIL-related metrics include an intra-tumor TIL% (iTIL%).
- the iTIL% is calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL.
- the TIL% value is a stromal TIL% (sTIL%).
- the sTIL% value is calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL.
- the one or more neural network models can include a segmentation algorithm for cell segmentation, which can be referred to as a cell segmentation algorithm.
- the cell segmentation algorithm may input an image and perform segmentation of the cells.
- the segmented cells may be labeled based on the region segmentation output.
- the segmented cell is labeled as a TIL cell if the cell is in a TIL region.
- the cell segmentation algorithm is configured to calculate one or more features related to the cells, also referred to herein as cell -based features, contained in the image.
- the calculated cell-based features may include morphology, color (nuclear), color (extra-nuclear), texture (energy, correlation, contrast, homogeneity, dissimilarity, angular second moment), chromatin skeleton morphology and color, among many other features that can be extracted.
- the cell segmentation algorithm may be configured to generate a number of TIL cells or average number of TIL cells. In various embodiments, the cell segmentation algorithm may be configured to generate an average tumor cell size. In various embodiments, the cell segmentation algorithm may be configured to generate an average TIL cell size. In various embodiments, the cell segmentation algorithm may be configured to generate an average distance between TIL cells.
- the cell segmentation algorithm may be configured to generate an average color of TIL cells. In various embodiments, the cell segmentation algorithm may be configured to generate an average distance between TIL cells vs tumor cells.
- the TIL-related metrics is a predicted TIL cluster spread in the identified tumor region. In various embodiments, the TIL-related metrics comprise a total count of TILs in the identified tumor region. In various embodiments, the TIL-related metrics can include a distance between TIL cells vs. distance between tumor cells. In various embodiments, the TIL-related metrics can include a distance of TIL cells from tumor cells. In various embodiments, the TIL-related metrics can include an average color of TIL cells. In various embodiments, the TIL-related metrics can include an average size of TIL cells.
- the one or more neural network models can include a classification algorithm that generates a TMB indicator that is either a TMB positive classification or TMB negative classification for the tissue sample.
- the classification algorithm generates the TMB positive classification if the tissue sample has a TMB value that exceeds a pre-set TMB threshold value.
- the one or more neural network models can include a regression algorithm that generates a TMB indicator that is a quantitative TMB score.
- the one or more neural network models can include another classification algorithm that generates a PD-L1 expression level indicator that is either a PD-L1 expression positive classification or a PD-L1 expression level negative classification.
- the classification algorithm generates a PD- 1 expression level indicator that is either a PD-1 expression positive classification or a PD-1 expression level negative classification.
- the classification algorithm generates the PD-L1 expression positive classification if the tissue sample has a PD-L1 expression level value that excess a pre-set PD-L1 expression level value.
- the classification algorithm generates the PD- 1 expression positive classification if the tissue sample has a PD-1 expression level value that excess a pre-set PD-1 expression level value.
- the one or more neural network models can include another regression algorithm that generates a PD-L1 expression level indicator that is a quantitative PD-L1 score.
- the another regression algorithm generates a PD-1 expression level indicator that is a quantitative PD-1 score.
- the quantitative PD-L1 score is a tumor proportion score (TPS) or a combined positive score (CPS).
- the quantitative PD-1 score is a tumor proportion score (TPS) or a combined positive score (CPS).
- the immuno-oncology therapeutic is a check point inhibitor.
- the check point inhibitor is a PD-L1 inhibitor or PD-1 inhibitor.
- the tissue sample is from bladder and the patient response to the immuno-oncology therapeutic is related to bladder cancer.
- a determination is made that the patient is or will be a responder based on the response status.
- the patient is administered a therapeutically effective amount of the immuno-oncology therapeutic.
- the tissue sample is from a patient known to have or suspected of having PD-L1 -positive cancer or PD-1 -positive cancer.
- the tissue sample is from a patient known to have or suspected of having non-small cell lung cancer, melanoma, Hodgkin lymphoma, bladder cancer, kidney cancer, or breast cancer.
- the immuno-oncology therapeutic targets PD-L1 or PD-1.
- the immuno-oncology therapeutic comprises one or more antibodies, adoptive cell therapies, immunomodulators, or a combination thereof.
- the antibody is a monoclonal antibody, a bispecific antibody, or a trispecific antibody.
- the adoptive cell therapy comprises immune cells that express one or more engineered antigen receptors.
- the engineered antigen receptor is a chimeric antigen receptor, a non-natural T cell receptor, or a combination thereof.
- the adoptive cell therapy comprises T cells, natural killer cells, natural killer T cells, or a combination thereof.
- the method SI 00 may optionally include the step of obtaining the sample from the patient. In various embodiments, the method SI 00 may include the step of diagnosing the patient as having cancer.
- FIG. 2C illustrates a method S300 for predicting a patient/treatment response to a therapeutic, in accordance with various embodiments.
- the method S300 can be implemented for example, via a computing system 300 as described with respect to Figure 3 below.
- the method S300 includes at step S310, obtaining or receiving, by one or more processors of the computing system 300, an image of a tissue sample, such as the image 105, from a patient; at step S320, generating, by the one or more processors, via a first neural network model, such as the model 120, one or more tumor mutation burden (TMB) indicators (e.g., as the output 130) from the one or more features; at step S330, generating, by the one or more processors, via a second neural network model, such as the model 140, one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death-1 (PD-1) expression level indicators (e.g., as the output 150) from
- TMB tumor
- the methods may lack the neural network model associated with one or more tumor mutation burden (TMB), as illustrated in Figures 2D-2F.
- TMB tumor mutation burden
- SI 10-2 corresponds to SI 10 of Figure 2A
- S120-2 corresponds to S130 of Figure 2A
- S130-2 corresponds to S140 of Figure 2A
- S140-2 corresponds to S150 of Figure 2A.
- S210-2 corresponds to S210 of Figure 2B.
- S220-2 corresponds to S220 of Figure 2B (except S220-2 lacks a step related to TMB indicators).
- S230-2 corresponds to S230 of Figure 2B (except S230-2 lacks a step related to TMB indicators).
- S240-2 corresponds to S240 of Figure 2B.
- S310-2 corresponds to S310 of Figure 2C.
- S320-2 corresponds to S330 of Figure 2C.
- S330-2 corresponds to S340 of Figure 2C.
- S340-2 corresponds to S350 of Figure 2C.
- S350-2 corresponds to S360 of Figure 2C (except S350-2 lacks a step related to TMB indicators.
- FIG. 3 is a block diagram illustrating a computer system 300 configured to perform a method of predicting a patient/treatment response to a therapeutic, with which embodiments of the disclosed systems and methods, or portions thereof may be implemented, in accordance with various embodiments.
- the illustrated computer system can be a local or remote computer system operatively connected to a control system for controlling or monitoring the systems and methods of the various embodiments herein.
- computer system 300 can include a bus 302 or other communication mechanism for communicating information and a processor 304 coupled with bus 302 for processing information.
- computer system 300 can also include a memory, which can be a random-access memory (RAM) 306 or other dynamic storage device, coupled to bus 302 for determining instructions to be executed by processor 304. Memory can also be used for storing temporary variables or other intermediate information during execution of instructions to be executed by processor 304.
- computer system 300 can further include a read only memory (ROM) 308 or other static storage device coupled to bus 302 for storing static information and instructions for processor 304.
- ROM read only memory
- a storage device 310 such as a magnetic disk or optical disk, can be provided and coupled to bus 302 for storing information and instructions.
- computer system 300 can be coupled via bus 302 to a display 312, such as a cathode ray tube (CRT) or liquid crystal display (LCD), for displaying information to a computer user.
- a display 312 such as a cathode ray tube (CRT) or liquid crystal display (LCD)
- An input device 314, including alphanumeric and other keys, can be coupled to bus 302 for communication of information and command selections to processor 304.
- a cursor control 316 such as a mouse, a trackball or cursor direction keys for communicating direction information and command selections to processor 304 and for controlling cursor movement on display 312.
- This input device 314 typically has two degrees of freedom in two axes, a first axis (i.e., x) and a second axis (i.e., y), that allows the device to specify positions in a plane.
- a first axis i.e., x
- a second axis i.e., y
- input devices 314 allowing for 3 -dimensional (x, y and z) cursor movement are also contemplated herein.
- components 312/314/316 together or individually, can make up a control system that connects the remaining components of the computer system to the systems herein and methods conducted on such systems, and controls execution of the methods and operation of the associated system.
- results can be provided by computer system 300 in response to processor 304 executing one or more sequences of one or more instructions contained in memory 306.
- Such instructions can be read into memory 306 from another computer-readable medium or computer-readable storage medium, such as storage device 310. Execution of the sequences of instructions contained in memory 306 can cause processor 304 to perform the processes described herein.
- hard-wired circuitry can be used in place of or in combination with software instructions to implement the present teachings.
- implementations of the present teachings are not limited to any specific combination of hardware circuitry and software.
- computer-readable medium e.g., data store, data storage, etc.
- computer-readable storage medium refers to any media that participates in providing instructions to processor 304 for execution.
- Such a medium can take many forms, including but not limited to, non-volatile media, volatile media, and transmission media.
- non-volatile media can include, but are not limited to, dynamic memory, such as memory 306.
- transmission media can include, but are not limited to, coaxial cables, copper wire, and fiber optics, including the wires that comprise bus 302.
- Computer-readable media include, for example, a floppy disk, a flexible disk, hard disk, magnetic tape, or any other magnetic medium, a CD-ROM, any other optical medium, punch cards, paper tape, any other physical medium with patterns of holes, a RAM, PROM, and EPROM, a FLASH-EPROM, another memory chip or cartridge, or any other tangible medium from which a computer can read.
- instructions or data can be provided as signals on transmission media included in a communications apparatus or system to provide sequences of one or more instructions to processor 304 of computer system 300 for execution.
- a communication apparatus may include a transceiver having signals indicative of instructions and data.
- the instructions and data are configured to cause one or more processors to implement the functions outlined in the disclosure herein.
- Representative examples of data communications transmission connections can include, but are not limited to, telephone modem connections, wide area networks (WAN), local area networks (LAN), infrared data connections, NFC connections, etc.
- computer system 300 can be implemented using computer system 300 as a standalone device or on a distributed network or shared computer processing resources such as a cloud computing network.
- the computer system 300 may include a single computer (or computer system) or multiple computers in communication with each other in a distributed implementation.
- the methodologies described herein may be implemented by various means depending upon the application. For example, these methodologies may be implemented in hardware, firmware, software, or any combination thereof.
- the processing unit may be implemented within one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, micro-controllers, microprocessors, electronic devices, other electronic units designed to perform the functions described herein, or a combination thereof.
- ASICs application specific integrated circuits
- DSPs digital signal processors
- DSPDs digital signal processing devices
- PLDs programmable logic devices
- FPGAs field programmable gate arrays
- processors controllers, micro-controllers, microprocessors, electronic devices, other electronic units designed to perform the functions described herein, or a combination thereof.
- the methods of the present teachings may be implemented as firmware and/or a software program and applications written in conventional programming languages such as C, C++, Python, etc. If implemented as firmware and/or software, the embodiments described herein can be implemented on a non-transitory computer-readable medium in which a program is stored for causing a computer to perform the methods described above. It should be understood that the various engines described herein can be provided on a computer system, such as computer system 300, whereby processor 304 would execute the analyses and determinations provided by these engines, subject to instructions provided by any one of, or a combination of, memory components 306/308/310 and user input provided via input device 314.
- TLB Tumor Mutation Burden
- the systems, compositions, and methods of the disclosure utilize tumor mutation burden (TMB) as at least one parameter or metric to determine whether or not a patient will be a responder to a particular type of therapy.
- the tumor mutation burden may include the total number of mutations (changes) present in the DNA of cancer cells.
- cancers including solid tumors
- TMB is defined as the number of non-inherited mutations per million bases (Mb) of investigated genomic sequence (Merino, et al. (2020). J Immunother Cancer.
- TMB tumor mutation burden
- a neural network model is configured to generate one or more TMB indicators from one or more features of a patient tissue sample.
- the neural network model comprises a classification algorithm, and the one or more TMB indicators comprises a TMB indicator for the tissue sample that is either a TMB positive classification or TMB negative classification.
- the neural network model comprises a classification algorithm, and the one or more TMB indicators comprises a TMB indicator for the tissue sample that is a positive classification if the tissue sample has a TMB value that exceeds a pre-set TMB threshold value.
- the output may be considered TMB High or TMB Low.
- the associated neural network model comprises a regression algorithm
- the one or more TMB indicators comprises a TMB indicator that is a quantitative TMB score.
- a TMB score may range from 0-100
- the pre-set TMB threshold value may be any value therein, including 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35,
- the TMB prediction parameter for the model is trained using hematoxylin and eosin (H&E) images, e.g., from muscle invasive bladder cancer (MIBC) patients.
- H&E hematoxylin and eosin
- MIBC muscle invasive bladder cancer
- the systems, methods, and compositions utilize checkpoint expression level as at least one parameter or metric to determine whether or not a patient will be a responder to a particular type of therapy.
- the checkpoint is any kind of checkpoint, in specific embodiments the checkpoint is an inhibitory checkpoint.
- the checkpoint is Programmed Death- 1 (PD-1), Programmed Death Ligand- 1 (PD-L1).
- the expression of the checkpoint is one that is quantified based on images of tissue samples from a patient.
- the expression of the checkpoint is one that is quantified based on H&E images of tissue samples from a patient.
- a neural network model may be trained based on one or more features from H&E images (e.g., from bladder cancer patients, including muscle-invasive bladder cancer patients) with respect to one or more PD-L1 expression level indicators or PD-1 expression level indicators.
- a neural network model comprises a classification algorithm and the one or more PD-L1 expression level indicators comprises a PD-L1 expression level indicator that is either a PD-L1 expression positive classification or a PD-L1 expression level negative classification.
- a neural network model comprises a classification algorithm and the one or more PD-1 expression level indicators comprises a PD-1 expression level indicator that is either a PD-1 expression positive classification or a PD-1 expression level negative classification.
- the classification algorithm generates the PD-L1 expression positive classification if the tissue sample has a PD-L1 expression level value that exceeds a pre-set PD-L1 expression level value.
- an output may be PD-L1 High or PD-L1 Low.
- the classification algorithm generates the PD-1 expression positive classification if the tissue sample has a PD-1 expression level value that exceeds a pre-set PD-1 expression level value.
- an output may be PD-1 High or PD-1 Low.
- the threshold for PD-L1 High and PD-L1 Low, or PD-1 High and PD-1 Low, may be different depending on the immunooncology therapeutic and/or the cancer. In certain embodiments, there are different thresholds that are applied based on the drug and cancer.
- a neural network model comprises a regression algorithm and the one or more PD-L1 expression level indicators comprises a quantitative PD-L1 score that may be a tumor proportion score (TPS) or a combined positive score (CPS).
- a neural network model comprises a regression algorithm and the one or more PD-1 expression level indicators comprises a quantitative PD-1 score that may be a tumor proportion score (TPS) or a combined positive score (CPS).
- TPS tumor proportion score
- CPS combined positive score
- the TPS score may range from 0-100, such as 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34,
- the CPS score may range from 0-100, such as 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58,
- PD-L1 High or PD-1 High may be a TPS or a CPS > 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14,
- a model directly predicts the TPS and/or the CPS score (i.e.. regression model).
- a model directly predicts PD-L1 High vs. PD-L1 Low for different thresholds, or PD-1 High vs. PD-1 Low for different thresholds.
- TIL Tumor Infiltrating Lymphocyte
- Tumor infiltrating lymphocytes are atype of white blood cell that has migrated from the blood towards a tumor. They can include both include T cells and B cells, and they recognize and kill cancer cells. In particular embodiments, the presence of lymphocytes in tumors is often associated with better clinical outcomes.
- the present disclosure provides systems, methods, and compositions that utilize TIL-related metrics as at least one parameter to determine whether or not a patient will be a responder to a particular type of therapy. The presence of TIL in a tumor sample is measured by any suitable method, including immunohistochemical staining, H&E staining, or both, in specific embodiments.
- the presence of TILs in the tumor is determined by a pathologist, and the pathologist may utilize IHC staining on tissue sample for visualization of TIL cells or they may be visualized in H&E slides. TILs have discerning characteristics to allow the pathologist to distinguish them from other cells, such as being more purple, darker, and rounder than tumor cells.
- TIL pathologist annotations on H&E images are utilized to train an algorithm.
- the pathologist assesses TILs in a stromal compartment of the tumor.
- the pathologist assesses TILs at the borders of an invasive tumor, for example. TILs outside of tumor borders, in zones of necrosis, in zones of fibrosis and/or associated with abscess are excluded, in specific cases. The pathologist may evaluate one or more H&E stained sections for the tumor.
- a neural network model is configured to calculate one or more TIL-related metrics from one or more features of a tissue sample.
- TIL-related metrics based on a TIL mask, tumor mask, necrosis mask, stroma masks, and cell nuclei segmentation mask.
- Several masks may be used to calculate “TIL-related” metrics (such as iTIL%, sTIL%, TIL clusters, etc.).
- the neural network may predict segmentation masks (i.e.. a TIL mask on top of the tissue image, along with other masks such as tumor, stroma, necrosis, etc., and cell nuclei segmentations as well).
- TIL tumor infiltrating lymphocyte
- the neural network model can be a cell segmentation algorithm that may input an image and perform segmentation of the cells.
- the segmented cells may be labeled based on the region segmentation output. For example, the segmented cell is labeled as a TIL cell if the cell is in a TIL region.
- the cell segmentation algorithm can calculate one or more cell-based features.
- the cell-based features may include morphology, color (nuclear), color (extra-nuclear), texture (energy, correlation, contrast, homogeneity, dissimilarity, angular second moment), chromatin skeleton morphology and color, among many other features that can be extracted.
- the cell segmentation algorithm may be configured to generate a number of TIL cells or average number of TIL cells, an average tumor cell size, an average TIL cell size, an average distance between TIL cells, an average color of TIL cells, an average distance between TIL cells vs tumor cells.
- the TIL-related metrics is a predicted TIL cluster spread in the identified tumor region, a total count of TILs in the identified tumor region, va distance between TIL cells vs. distance between tumor cells, a distance of TIL cells from tumor cells, an average color of TIL cells, and/or an average size of TIL cells.
- the present disclosure provides embodiments wherein an individual in need of immune-oncology therapy is evaluated for being a responder to the immune -oncology therapy.
- Embodiments of the disclosure include predicting patient response to an immuno-oncology therapeutic.
- the immuno-oncology therapeutic is an immune check point inhibitor.
- the immuno-oncology therapeutic may be an antibody or may be adoptive cell therapy with immune cells, for example, or both may be utilized.
- the antibody is a monoclonal antibody, and the monoclonal antibody may target any checkpoint.
- the antibody may be monospecific or bispecific.
- the antibody may target PD-L1 or PD-1, for example.
- the cells may be immune cells, such as T cells, natural killer (NK) cells, or NK T cells, as examples.
- the cells may be engineered to express a non-natural protein, such as an antigen receptor, and the antigen receptor may comprise an extracellular domain that is an antibody that targets an immune checkpoint.
- the antigen receptor comprises an extracellular domain that binds an immune checkpoint.
- the immune- oncologic therapeutic is a check point inhibitor such as a PD-L1 inhibitor or PD-1 inhibitor.
- one may combine the immuno-oncology therapeutic with one or more other types of cancer therapies, such as surgery, radiation, hormone therapy, other immunotherapies, or chemotherapy.
- methods comprise utilizing an image of a tissue sample from a patient, wherein the tissue sample comprises one or more features.
- the method comprises obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample.
- the method may comprise generating, such as by the one or more processors, via one or more neural network models: (a) one or more tumor mutation burden (TMB) indicators from the one or more features; (b) one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death-1 (PD-1) expression level indicators from the one or more features; and (c) one or more tumor infiltrating lymphocyte (TIL) metric values from the one or more features; predicting, by the one or more processors, via the one or more neural network models, a response status of the patient to an immuno-oncology therapeutic using the one or more TMB indicators, the one or more PD-L1 expression level indicators or the one or more PD- 1 expression level indicators, respectively, and the one or more TIL metrics; and selecting, based on the response status of the patient, whether to include the patient in a cohort of patients participating in a clinical trial or whether to administer to the patient a therapeutically effective amount of one or more immune
- the method comprises utilizing an image of a tissue sample from a patient, wherein the tissue sample comprises one or more features; obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; generating, by the one or more processors, via one or more neural network models: (a) one or more tumor mutation burden (TMB) indicators from the one or more features; (b) one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death-1 (PD-1) expression level indicators from the one or more features; and (c) one or more tumor infiltrating lymphocyte (TIL) percentage (%) values (as an example of a TIL metric values) from the one or more features; predicting, by the one or more processors, via the one or more neural network models, a response status of the patient to an immuno-oncology therapeutic using the one or more TMB indicators, the one or more PD
- TMB tumor mutation burden
- Various embodiments of the method include obtaining, by one or more processors, an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample.
- the party that obtains the image is the same party that handles the processing and/or generating steps. In other cases, the party that obtains the image is not the same party that handles the processing and/or generating steps.
- the tissue sample may be obtained from storage or may be evaluated soon after collection.
- the tissue sample may or may not originate from a biopsy of an individual undergoing diagnosis of cancer. The individual may be known to have cancer or may be suspected of having cancer, based on one or more symptoms.
- the systems, methods, and compositions of the disclosure may be utilized for any type of cancer, including hematological and solid tumors.
- the cancer may be of any stage or grade; it may or may not be refractory; it may or may not be metastatic.
- the cancer may be of the bladder, brain, breast, lung, colon, pancreas, prostate, liver, spleen, kidney, stomach, cervix, uterus, ovary, testes, gall bladder, bone, thyroid, skin, endometrium, rectum, and so forth.
- the subject has one or more symptoms of bladder cancer, such as hematuria; having to urinate more often than usual; pain or burning during urination; feeling a need to urinate right away, even when the bladder is not full; having trouble urinating or having a weak urine stream; having to get up to urinate many times during the night; being unable to urinate; lower back pain on one side; loss of appetite and weight loss; feeling tired or weak; swelling in the feet; bone pain; or a combination thereof.
- the subject may be suspected of having bladder cancer or may be known to have bladder cancer.
- Diagnosis of bladder cancer can occur using cystoscopy, biopsy, blood tests, urine cytology, computerized tomography (CT) urogram, retrograde pyelogram, or a combination thereof. Other tests may be run to determine the extent of the cancer, such as Magnetic resonance imaging (MRI), Positron emission tomography (PET), Bone scan, and/or Chest X-ray.
- the bladder cancer may be urothelial carcinoma, squamous cell carcinoma, or adenocarcinoma.
- the bladder cancer may be noninvasive, non-muscle-invasive, or muscle-invasive.
- the bladder cancer may be of any stage or grade, including noninvasive papillary carcinoma, carcinoma in situ, Tl, T2, T3, or T4.
- Embodiments of the disclosure include methods of treatment for individuals determined to be responders based on systems, methods, and compositions encompassed herein.
- a responder individual is administered a therapeutically effective amount of one or more immune-oncology therapeutics based on measuring performed by systems, methods, and compositions encompassed herein.
- an individual that is not a responder is not administered one or more immune-oncology therapeutics based on measuring performed by systems, methods, and compositions encompassed herein.
- a therapeutically effective amount of one or more immuno-oncology therapeutics is administered to an individual determined to have, or predicted to have, a response to one or more immuno-oncology therapeutics based on having the appropriate one or more TMB indicators, one or more PD-L1 and/or PD-1 expression level indicators, and one or more TIL-related metrics, all from one or more features of an image of a tissue sample.
- a therapeutically effective amount of one or more immuno-oncology therapeutics is administered to an individual determined to have, or predicted to have, a response to one or more immuno-oncology therapeutics based on the following: providing an image of a tissue sample from a patient, the tissue sample comprising and one or more features of the tissue sample; generating, by one or more processors, via a first neural network model, a one or more TMB indicators from the one or more features; generating, by the one or more processors, via a second neural network model, a one or more PD-L1 expression level indicators from the one or more features; generating, by the one or more processors, via a third neural network model, one or more TIL-related metrics from the one or more features; and predicting, by the one or more processors, via a fourth neural network model, that the individual will respond to the immuno-oncology therapeutic using the one or more TMB indicators, the one or more PD-L1 expression level indicators, and the one or more TIL-related metrics
- a therapeutically effective amount of one or more immuno-oncology therapeutics is administered to an individual determined to have, or predicted to have, a response to one or more immuno-oncology therapeutics based on the following: obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; generating, by the one or more processors, via one or more neural network models: (a) one or more tumor mutation burden (TMB) indicators from the one or more features; (b) one or more programmed deathligand 1 (PD-L1) expression level indicators from the one or more features; and (c) one or more tumor infiltrating lymphocyte (TIL) percentage (%) values from the one or more features; determining or predicting, by the one or more processors, via the one or more neural network models, that the patient will respond to an immuno-oncology therapeutic using the one or more TMB indicators, the one or more PD-L1 expression level indicators, and the one or more TIL- related
- Embodiment 1 A system for predicting patient response to an immuno-oncology therapeutic, the system comprising: a data store for storing an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample; a computing device communicatively connected to the data store, the computing device comprising: a first neural network model configured to generate one or more tumor mutation burden (TMB) indicators from the one or more features; a second neural network model configured to generate one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or configured to generate one or more programmed death- 1 (PD-1) expression level indicators from the one or more features; a third neural network model configured to calculate one or more tumor infiltrating lymphocyte (TIL)-related metrics from the one or more features; or a fourth neural network model configured to predict a response status of the patient to the immuno-oncology therapeutic in accordance with the one or more TMB indicators, the one or more PD-L1 expression level indicators or the one or more
- Embodiment 2 The system of embodiment 1, wherein the prediction of the response status of the patient to the immuno-oncology therapeutic is performed in accordance with one or more medical history features of the patient.
- Embodiment 3 The system of embodiment 1 or 2, wherein the third neural network model comprises a segmentation algorithm configured to generate an identification of a tumor region in the tissue sample based on the one or more features, wherein the identification is used to further generate the one or more TIL-related metrics.
- the third neural network model comprises a segmentation algorithm configured to generate an identification of a tumor region in the tissue sample based on the one or more features, wherein the identification is used to further generate the one or more TIL-related metrics.
- Embodiment 4 The system of embodiment 3, wherein the segmentation algorithm is configured to predict a TIL mask, which is used in combination with a tumor mask to calculate one or more TIL related metrics.
- Embodiment s The system of embodiment 4, wherein the one or more TIL- related metrics comprises a predicted number of TIL clusters in the identified tumor region.
- Embodiment 6 The system of embodiment 4, wherein the one or more TIL- related metrics comprises a predicted size of TIL clusters in the identified tumor region.
- Embodiment 7 The system of embodiment 4, wherein the one or more TIL- related metrics comprises a predicted TIL cluster spread in the identified tumor region.
- Embodiment 8 The system of embodiment 4, wherein the one or more TIL- related metrics comprises a total count of TILs in the identified tumor region.
- Embodiment 9 The system of embodiment 4, wherein the one or more TIL- related metrics comprises distance between TIL cells vs. distance between tumor cells
- Embodiment 10 The system of embodiment 4, wherein the one or more TIL- related metrics comprises distance of TIL cells from tumor cells.
- Embodiment 11 The system of embodiment 4, wherein the one or more TIL- related metrics comprises average color of TIL cells.
- Embodiment 12 The system of embodiment 4, wherein the one or more TIL- related metrics comprises average size of TIL cells.
- Embodiment 13 The system of any one of embodiments 1-12, wherein the one or more TIL-related metrics comprises an intra-tumor TIL% (iTIL%).
- iTIL% intra-tumor TIL%
- Embodiment 14 The system of embodiment 13, wherein the iTIL% is calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL.
- Embodiment 15 The system of embodiment 1, wherein the one or more TIL- related metrics comprises a stromal TIL% (sTIL%).
- sTIL% stromal TIL%
- Embodiment 16 The system of embodiment 15, wherein the sTIL% value is calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL.
- Embodiment 17 The system of any one of embodiments 1-16, wherein the first neural network model comprises a classification algorithm and the one or more TMB indicators comprises a TMB indicator for the tissue sample that is either a TMB positive classification or TMB negative classification.
- Embodiment 18 The system of embodiment 17, wherein the classification algorithm generates the TMB positive classification if the tissue sample has a TMB value that exceeds a pre-set TMB threshold value.
- Embodiment 19 The system of any one of embodiments 1-16, wherein the first neural network model comprises a regression algorithm and the one or more TMB indicators comprises a TMB indicator that is a quantitative TMB score.
- Embodiment 20 The system of any one of embodiments 1-19, wherein the second neural network model comprises a classification algorithm and the one or more PD-L1 expression level indicators comprises a PD-L1 expression level indicator that is either a PD-L1 expression positive classification or a PD-L1 expression level negative classification.
- Embodiment 21 The system of any one of embodiments 1-19, wherein the second neural network model comprises a classification algorithm and the one or more PD-1 expression level indicators comprises a PD-1 expression level indicator that is either a PD- 1 expression positive classification or a PD-1 expression level negative classification.
- Embodiment 22 The system of embodiment 20, wherein the classification algorithm generates the PD-L1 expression positive classification if the tissue sample has a PD-L1 expression level value that exceeds a pre-set PD-L1 expression level value.
- Embodiment 23 The system of embodiment 21, wherein the classification algorithm generates the PD-1 expression positive classification if the tissue sample has a PD-1 expression level value that exceeds a pre-set PD-1 expression level value.
- Embodiment 24 The system of any one of embodiments 1-19, wherein the second neural network model comprises a regression algorithm and the one or more PD-L1 expression level indicators comprises a quantitative PD-L1 score.
- Embodiment 25 The system of embodiment 1 , wherein the second neural network model comprises a regression algorithm and the one or more PD-1 expression level indicators comprises a quantitative PD-1 score.
- Embodiment 26 The system of embodiment 24, wherein the quantitative PD-L1 score is tumor proportion score (TPS).
- TPS tumor proportion score
- Embodiment 27 The system of embodiment 24, wherein the quantitative PD-L1 score is a combined positive score (CPS).
- CPS combined positive score
- Embodiment 28 The system of embodiment 25, wherein the quantitative PD-1 score is tumor proportion score (TPS).
- TPS tumor proportion score
- Embodiment 29 The system of embodiment 25, wherein the quantitative PD-1 score is a combined positive score (CPS).
- CPS combined positive score
- Embodiment 30 The system of any one of embodiments 1-29, wherein the immuno-oncology therapeutic is a check point inhibitor.
- Embodiment 31 The system of embodiment 30, wherein the check point inhibitor is a PD-L1 inhibitor or PD-1 inhibitor.
- Embodiment 32 The system of any one of embodiments 1-31, wherein the third neural network model is further configured to generate one or more TIL masks from the one or more features; and wherein the fourth neural network model is further configured to predict the response status of the patient to the immuno-oncology therapeutic using the one or more TIL masks.
- Embodiment 33 The system of any one of embodiments 1, wherein the tissue sample is from the patient’s bladder; and wherein the patient response to the immuno- oncology therapeutic is related to bladder cancer.
- Embodiment 34 The system of any one of embodiments 1-33, wherein the tissue sample is from a patient known to have PD-L1 -positive cancer.
- Embodiment 35 The system of any one of embodiments 1-33, wherein the tissue sample is from a patient known to have PD-1 -positive cancer.
- Embodiment 36 The system of any one of embodiments 1-35, wherein the tissue sample is from a patient known to have any one or more of the following: non-small cell lung cancer, melanoma, Hodgkin lymphoma, bladder cancer, kidney cancer, and breast cancer.
- Embodiment 37 The system of any one of embodiments 1-36, wherein the immuno-oncology therapeutic targets PD-L1 or PD-1.
- Embodiment 38 The system of any one of embodiments 1-37, wherein the immuno-oncology therapeutic comprises one or more antibodies, one or more adoptive cell therapies, or a combination thereof.
- Embodiment 39 The system of embodiment 38, wherein the one or more antibodies comprises at least one of: a monoclonal antibody, a bispecific antibody, or a trispecific antibody.
- Embodiment 40 The system of embodiment 38, wherein the one or more adoptive cell therapies comprises immune cells that express one or more engineered antigen receptors.
- Embodiment 41 The system of embodiment 40, wherein the one or more engineered antigen receptors comprises at least one of: a chimeric antigen receptor, a nonnatural T cell receptor, or a combination thereof.
- Embodiment 42 The system of any one of embodiments 38-41, wherein the one or more adoptive cell therapies comprises at least one of: T cells, natural killer cells, natural killer T cells, or a combination thereof.
- Embodiment 43 The system of embodiment 1 or 2, wherein the third neural network model comprises a region/TIL segmentation algorithm and a cell segmentation algorithm.
- Embodiment 44 The system of embodiment 43, wherein the cell segmentation algorithm calculates one or more cell -based features.
- Embodiment 45 The system of embodiment 44, wherein the one or more cellbased features include morphology, color (nuclear), color (extra-nuclear), texture (energy, correlation, contrast, homogeneity, dissimilarity, or angular second moment).
- Embodiment 46 The system of embodiment 43, wherein the cell segmentation algorithm generates a number of TIL cells or average number of TIL cells.
- Embodiment 47 The system of embodiment 43, wherein the cell segmentation algorithm generates an average tumor cell size.
- Embodiment 48 The system of embodiment 43, wherein the cell segmentation algorithm generates an average TIL cell size.
- Embodiment 49 The system of embodiment 43, wherein the cell segmentation algorithm generates an average distance between TIL cells.
- Embodiment 50 The system of embodiment 43, wherein the cell segmentation algorithm generates an average color of TIL cells.
- Embodiment 51 The system of embodiment 43, wherein the cell segmentation algorithm generates an average distance between TIL cells vs tumor cells.
- Embodiment 52 A method for predicting patient response to an immunooncology therapeutic, the method comprising: obtaining, by one or more processors, an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample; generating, by the one or more processors, via a first neural network model, one or more tumor mutation burden (TMB) indicators from the one or more features; generating, by the one or more processors, via a second neural network model, one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death-1 (PD-1) expression level indicators from the one or more features; generating, by the one or more processors, via a third neural network model, one or more tumor infiltrating lymphocyte (TIL) related metrics from the one or more features; and predicting, by the one or more processors, via a fourth neural network model, a response status of the patient to the immuno-oncology therapeutic using the one or more TMB indicators, the one
- TMB
- Embodiment 53 The method of embodiment 52, further comprising:
- Embodiment 54 The method of embodiment 52 or 53, further comprising: displaying, on a display screen, a graphical representation indicating the response status of the patient.
- Embodiment 55 The method of any one of embodiments 52-54, wherein the prediction of the response status of the patient to the immuno-oncology therapeutic further uses one or more medical history features of the patient.
- Embodiment 56 The method of any one of embodiments 52-55, wherein the third neural network is a segmentation algorithm that identifies a tumor region in the tissue sample.
- Embodiment 57 The method of embodiment 56, wherein the segmentation algorithm predicts one or more TIL mask structure features in the identified tumor region to generate the one or more TIL-related metrics.
- Embodiment 58 The method of embodiment 57, wherein the predicted TIL mask structure feature is a predicted number of TIL clusters in the identified tumor region.
- Embodiment 59 The method of embodiment 57, wherein the predicted TIL mask structure feature is a predicted size of TIL clusters in the identified tumor region.
- Embodiment 60 The method of cla embodiment im 57, wherein the predicted TIL mask structure feature is a predicted TIL cluster spread in the identified tumor region.
- Embodiment 61 The method of embodiment 57, wherein the one or more TIL mask structure features comprises a total count of TILs in the identified tumor region.
- Embodiment 62 The method of embodiment 57, wherein the one or more TIL mask structure features comprises distance between TIL cells vs. distance between tumor cells
- Embodiment 63 The method of embodiment 57, wherein the one or more TIL mask structure features comprises distance of TIL cells from tumor cells.
- Embodiment 64 The method of embodiment 57, wherein the one or more TIL mask structure features comprises average color of TIL cells.
- Embodiment 65 The method of embodiment 57, wherein the one or more TIL mask structure features comprises average size of TIL cells.
- Embodiment 66 The method of any one of embodiments 52-65, wherein the TIL% value is an intra-tumor TIL% (iTIL%).
- iTIL% intra-tumor TIL%
- Embodiment 67 The method of embodiment 66, wherein the iTIL% is calculated as a percent of the identified tumor region in the tissue sample that is infiltrated by TIL.
- Embodiment 68 The method of any one of embodiments 52-65, wherein the TIL% value is a stromal TIL% (sTIL%).
- Embodiment 69 The method of embodiment 68, wherein the sTIL% value is calculated as a percent of stroma in the identified tumor region in the tissue sample that is infiltrated by TIL.
- Embodiment 70 The method of any one of embodiments 52-69, wherein the first neural network model is a classification algorithm that generates a TMB indicator that is either a TMB positive classification or TMB negative classification for the tissue sample.
- Embodiment 71 The method of embodiment 70, wherein the classification algorithm generates the TMB positive classification if the tissue sample has a TMB value that exceeds a pre-set TMB threshold value.
- Embodiment 72 The method of any one of embodiments 52-69, wherein the first neural network model is a regression algorithm that generates a TMB indicator that is a quantitative TMB score.
- Embodiment 73 The method of any one of embodiments 52-72, wherein the second neural network model is a classification algorithm that generates a PD-L1 expression level indicator that is either a PD-L1 expression positive classification or a PD- L1 expression level negative classification.
- Embodiment 74 The method of any one of embodiments 52-72, wherein the second neural network model is a classification algorithm that generates a PD-1 expression level indicator that is either a PD-1 expression positive classification or a PD-1 expression level negative classification.
- Embodiment 75 The method of embodiment 73, wherein the classification algorithm generates the PD-L1 expression positive classification if the tissue sample has a PD-L1 expression level value that excess a pre-set PD-L1 expression level value.
- Embodiment 76 The method of embodiment 74, wherein the classification algorithm generates the PD-1 expression positive classification if the tissue sample has a PD-1 expression level value that excess a pre-set PD-1 expression level value.
- Embodiment 77 The method of any one of embodiments 52-72, wherein the second neural network model is a regression algorithm that generates a PD-L1 expression level indicator that is a quantitative PD-L1 score.
- Embodiment 78 The method of any one of embodiments 52-72, wherein the second neural network model is a regression algorithm that generates a PD-1 expression level indicator that is a quantitative PD-1 score.
- Embodiment 79 The method of embodiment 77, wherein the quantitative PD-L1 score is tumor proportion score (TPS).
- TPS tumor proportion score
- Embodiment 80 The method of embodiment 78, wherein the quantitative PD-1 score is tumor proportion score (TPS).
- TPS tumor proportion score
- Embodiment 81 The method of embodiment 77, wherein the quantitative PD-L1 score is a combined positive score (CPS).
- Embodiment 82 The method of embodiment 78, wherein the quantitative PD-1 score is a combined positive score (CPS).
- CPS combined positive score
- Embodiment 83 The method of any one of embodiments 52-82, wherein the immuno-oncology therapeutic is a check point inhibitor.
- Embodiment 84 The method of embodiment 83 , wherein the check point inhibitor is a PD-L1 inhibitor or PD-1 inhibitor.
- Embodiment 85 The method of any one of embodiments 52-84, further comprising: generating, by the one or more processors, via the third neural network model, one or more TIL masks from the one or more features, wherein the predicting of the response status of the patient to the immuno-oncology therapeutic further comprises using the one or more TIL masks.
- Embodiment 86 The method of any one of embodiments 52-85, wherein the tissue sample is from bladder and the patient response to the immuno-oncology therapeutic is related to bladder cancer.
- Embodiment 87 The method of any one of embodiments 52-86, wherein a determination is made that the patient is or will be a responder based on the response status; and wherein, in accordance with the determination, the patient is administered a therapeutically effective amount of the immuno-oncology therapeutic.
- Embodiment 88 The method of any one of embodiments 52-86, wherein a determination is made that the patient is not or will not be a responder based on the response status; and wherein, in accordance with the determination, the patient is not administered the immuno-oncology therapeutic.
- Embodiment 89 The method of any one of embodiments 52-88, wherein the tissue sample is from a patient known to have PD-L1 -positive cancer or PD-1 -positive cancer.
- Embodiment 90 The method of any one of embodiments 52-89, wherein the tissue sample is from a patient known to have any of the following: non-small cell lung cancer, melanoma, Hodgkin lymphoma, bladder cancer, kidney cancer, or breast cancer.
- Embodiment 91 The method of any one of embodiments 52-90, wherein the immuno-oncology therapeutic targets PD-L1 or PD-1.
- Embodiment 92 The method of any one of embodiments 52-91, wherein the immuno-oncology therapeutic comprises one or more antibodies, adoptive cell therapies, immunomodulators, or a combination thereof.
- Embodiment 93 The method of embodiment 92, wherein the antibody is a monoclonal antibody, a bispecific antibody, or a trispecific antibody.
- Embodiment 94 The method of embodiment 92, wherein the adoptive cell therapy comprises immune cells that express one or more engineered antigen receptors.
- Embodiment 95 The method of embodiment 94, wherein the engineered antigen receptor is a chimeric antigen receptor, a non-natural T cell receptor, or a combination thereof.
- Embodiment 96 The method of any one of embodiments 92-94, wherein the adoptive cell therapy comprises T cells, natural killer cells, natural killer T cells, or a combination thereof.
- Embodiment 97 The method of any one of embodiments 52-96, further comprising: obtaining the sample from the patient.
- Embodiment 98 The method of any one of embodiments 52-97, further comprising: diagnosing the patient as having cancer.
- Embodiment 99 A method comprising: obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; generating, by the one or more processors, via one or more neural network models one or more of the following: (a) one or more tumor mutation burden (TMB) indicators from the one or more features; (b) one or more programmed death-ligand 1 (PD-L1) expression level indicators from the one or more features or one or more programmed death-1 (PD-1) expression level indicators from the one or more features; and (c) one or more tumor infdtrating lymphocyte (TIL) percentage (%) values from the one or more features; predicting, by the one or more processors, via the one or more neural network models, a response status of the patient to an immuno-oncology therapeutic using one or more of: the one or more TMB indicators, the one or more PD-L1 expression level indicators, the one or more PD-1 expression level indicators, and/or the one or
- TMB
- Embodiment 100 A method comprising: obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; and generating, by the one or more processors, via one or more neural network models a prediction of a response status of the patient to an immuno-oncology therapeutic using the image.
- Embodiment 102 A method comprising: obtaining, by one or more processors from a data source, an image of a tissue sample from a patient and one or more features of the tissue sample; and training, by the one or more processors, one or more neural network models to predict a response status of the patient to an immuno-oncology therapeutic using the image.
- Embodiment 103 A method for predicting patient response to an immuno- oncology therapeutic, the method comprising: obtaining, by one or more processors, an image of a tissue sample from a patient, the image comprising one or more features of the tissue sample; generating, by the one or more processors, one or more tumor mutation burden (TMB) indicators from the one or more features; generating, by the one or more processors, from the one or more features, one or more programmed death-ligand 1 (PD- Ll) expression level indicators or one or more programmed death-1 (PD-1) expression level indicators; generating, by the one or more processors, one or more tumor infiltrating lymphocyte (TIL) related metrics from the one or more features; and training, by the one or more processors, a fourth neural network model to predict a response status of the patient to the immuno-oncology therapeutic based on one or more of: the one or more TMB indicators, the one or more PD-L1 expression level indicators, the one or more PD-1 level indicators
- Embodiment 104 A method of treating a subject suffering from bladder cancer, the method comprising: administering to the subject an immuno-oncology therapeutic, wherein the subject has been determined to be responsive to the immuno-oncology therapeutic via a trained machine learning classifier that distinguishes between responsive and non- responsive subjects who have received the immuno-oncology therapeutic, based at least in part on analyzing in the subject of one or more of: a PD-L1 expression level, a PD-1 expression level, and one or more tumor infiltrating lymphocyte (TIL)-related metrics.
- TIL tumor infiltrating lymphocyte
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Epidemiology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Bioethics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Public Health (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biotechnology (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Theoretical Computer Science (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
Sont divulgués des systèmes et des procédés de prédiction de la réponse d'un patient à un agent thérapeutique. Les systèmes et les procédés peuvent comprendre une mémoire de données servant à stocker une image d'un échantillon de tissu provenant d'un patient, l'image comprenant une ou plusieurs caractéristiques de l'échantillon de tissu ; un ordinateur comprenant : un premier modèle de réseau neuronal conçu pour générer un ou plusieurs indicateurs du niveau d'expression de ligands de mort cellulaire programmée 1 (PD-L1) ou conçu pour générer un ou plusieurs indicateurs du niveau d'expression de la protéine de la mort cellulaire programmée 1 (PD-1) ; un deuxième modèle de réseau neuronal conçu pour calculer un ou plusieurs paramètres associés aux lymphocytes infiltrant les tumeurs (TIL) ; et/ou un troisième modèle de réseau neuronal conçu pour prédire un état de réponse du patient à l'agent thérapeutique immuno-oncologique conformément audit ou auxdits indicateurs de niveau d'expression de PD-L1 et/ou audit ou auxdits indicateurs de niveau d'expression de PD-1, respectivement, et/ou audit ou auxdits paramètres associés aux TIL ; et un système d'affichage conçu pour afficher un résultat qui indique l'état de réponse du patient.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US202263403652P | 2022-09-02 | 2022-09-02 | |
US63/403,652 | 2022-09-02 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2024047608A1 true WO2024047608A1 (fr) | 2024-03-07 |
Family
ID=88017948
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/IB2023/058673 WO2024047608A1 (fr) | 2022-09-02 | 2023-09-01 | Procédés et systèmes de profilage immunitaire basé sur l'ia pour la stratification de patients cancéreux |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2024047608A1 (fr) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CA3133826A1 (fr) * | 2019-03-26 | 2020-10-01 | Tempus Labs, Inc. | Determination de biomarqueurs a partir d'images de lames histopathologiques |
US20220101519A1 (en) * | 2018-05-14 | 2022-03-31 | Tempus Labs, Inc. | Determining Biomarkers from Histopathology Slide Images |
US20220154284A1 (en) * | 2020-11-19 | 2022-05-19 | Tempus Labs, Inc. | Determination of cytotoxic gene signature and associated systems and methods for response prediction and treatment |
-
2023
- 2023-09-01 WO PCT/IB2023/058673 patent/WO2024047608A1/fr unknown
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20220101519A1 (en) * | 2018-05-14 | 2022-03-31 | Tempus Labs, Inc. | Determining Biomarkers from Histopathology Slide Images |
CA3133826A1 (fr) * | 2019-03-26 | 2020-10-01 | Tempus Labs, Inc. | Determination de biomarqueurs a partir d'images de lames histopathologiques |
US20220154284A1 (en) * | 2020-11-19 | 2022-05-19 | Tempus Labs, Inc. | Determination of cytotoxic gene signature and associated systems and methods for response prediction and treatment |
Non-Patent Citations (1)
Title |
---|
MERINO ET AL., 1 IMMUNOTHER CANCER, vol. 8, no. 1, 2020, pages e000147 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Faria et al. | Optimising the diagnosis of prostate cancer in the era of multiparametric magnetic resonance imaging: a cost-effectiveness analysis based on the prostate MR imaging study (PROMIS) | |
CN107492090B (zh) | 根据合成数据利用机器学习的基于图像的肿瘤表型分析 | |
Trebeschi et al. | Prognostic value of deep learning-mediated treatment monitoring in lung cancer patients receiving immunotherapy | |
US20100113299A1 (en) | Gene and gene expressed protein targets depicting biomarker patterns and signature sets by tumor type | |
EP2435940B1 (fr) | Procédé et dispositif de pronostic et de surveillance d'effets secondaires | |
US20240087121A1 (en) | Systems and methods to process electronic images for continuous biomarker prediction | |
US11308616B2 (en) | Systems and methods to process electronic images to provide image-based cell group targeting | |
KR102562708B1 (ko) | 일반화된 질병 검출을 위해 전자 이미지들을 처리하기 위한 시스템들 및 방법들 | |
EP3836157B1 (fr) | Procédé pour obtenir des informations cliniques liées à une maladie | |
WO2022020722A1 (fr) | Prédiction de réponse à un traitement d'immunothérapie en utilisant l'analyse par apprentissage profond de données d'imagerie et cliniques | |
Trebeschi et al. | Development of a prognostic AI-monitor for metastatic urothelial cancer patients receiving immunotherapy | |
Manuali et al. | A web-based tumor registration system for a regional Canine Cancer Registry in Umbria, central Italy | |
Bessen et al. | Does one size fit all? Cost utility analyses of alternative mammographic follow-up schedules, by risk of recurrence | |
WO2024047608A1 (fr) | Procédés et systèmes de profilage immunitaire basé sur l'ia pour la stratification de patients cancéreux | |
US20160224738A1 (en) | System, Method and Software for Predicting Drug Efficacy in a Patient | |
WO2023154573A1 (fr) | Techniques d'apprentissage automatique pour détection de structure lymphoïde tertiaire (tls) | |
Kronberg | Applications of Supervised Deep (Transfer) Learning for Medical Image Classification | |
US20220148731A1 (en) | Systems and Methods for Uncertainty Quantification in Radiogenomics | |
Moorthy et al. | Identification of Neuro-Blastoma and Pediatric Cancer Using Time Series Analysis | |
Sui et al. | Image Bio-Markers and Gene Expression Data Correlation Framework for Lung Cancer Radio-Genomics Analysis Based on Deep Learning | |
Marchevsky | The Use of Artificial Neural Networks for the Diagnosis and Estimation of Prognosis in Cancer Patients | |
Udager | Targeting FOXA1-Mediated Epigenetic Reprogramming in Aggressive Salivary Gland Cancer | |
Zhang et al. | A radiomics based approach using adrenal gland and periadrenal fat CT images to allocate COVID-19 health care resources fairly | |
Saednia | Machine Learning and Digital Histopathology Analysis for Tissue Characterization and Treatment Response Prediction in Breast Cancer | |
Ditsch et al. | Digital Medicine and Artificial Intelligence in the Area of Breast and Gynecologic Cancer |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23768624 Country of ref document: EP Kind code of ref document: A1 |