US20220341930A1 - Diagnostic to distinguish bacterial infections - Google Patents
Diagnostic to distinguish bacterial infections Download PDFInfo
- Publication number
- US20220341930A1 US20220341930A1 US17/806,242 US202217806242A US2022341930A1 US 20220341930 A1 US20220341930 A1 US 20220341930A1 US 202217806242 A US202217806242 A US 202217806242A US 2022341930 A1 US2022341930 A1 US 2022341930A1
- Authority
- US
- United States
- Prior art keywords
- bacterial
- peptides
- samples
- infections
- infection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
- 208000035143 Bacterial infection Diseases 0.000 title claims abstract description 48
- 208000022362 bacterial infectious disease Diseases 0.000 title claims abstract description 48
- 208000036142 Viral infection Diseases 0.000 claims abstract description 47
- 230000009385 viral infection Effects 0.000 claims abstract description 46
- 108090000765 processed proteins & peptides Proteins 0.000 claims description 96
- 102000004196 processed proteins & peptides Human genes 0.000 claims description 77
- 230000027455 binding Effects 0.000 claims description 10
- 230000004044 response Effects 0.000 claims description 4
- 238000000034 method Methods 0.000 abstract description 16
- 239000003242 anti bacterial agent Substances 0.000 abstract description 12
- 229940088710 antibiotic agent Drugs 0.000 abstract description 11
- 230000003115 biocidal effect Effects 0.000 abstract description 8
- 238000003491 array Methods 0.000 abstract description 4
- 238000003556 assay Methods 0.000 abstract description 4
- 230000001580 bacterial effect Effects 0.000 description 58
- 241000894006 Bacteria Species 0.000 description 40
- 238000012360 testing method Methods 0.000 description 32
- 208000015181 infectious disease Diseases 0.000 description 31
- 239000000523 sample Substances 0.000 description 29
- 241000700605 Viruses Species 0.000 description 20
- 238000010200 validation analysis Methods 0.000 description 19
- 210000000987 immune system Anatomy 0.000 description 17
- 238000012549 training Methods 0.000 description 17
- 238000002474 experimental method Methods 0.000 description 10
- 244000052769 pathogen Species 0.000 description 10
- 230000035945 sensitivity Effects 0.000 description 10
- 206010022000 influenza Diseases 0.000 description 9
- 238000003745 diagnosis Methods 0.000 description 8
- 230000003612 virological effect Effects 0.000 description 8
- 230000005875 antibody response Effects 0.000 description 7
- 108090000623 proteins and genes Proteins 0.000 description 7
- 210000002966 serum Anatomy 0.000 description 7
- FWMNVWWHGCHHJJ-SKKKGAJSSA-N 4-amino-1-[(2r)-6-amino-2-[[(2r)-2-[[(2r)-2-[[(2r)-2-amino-3-phenylpropanoyl]amino]-3-phenylpropanoyl]amino]-4-methylpentanoyl]amino]hexanoyl]piperidine-4-carboxylic acid Chemical compound C([C@H](C(=O)N[C@H](CC(C)C)C(=O)N[C@H](CCCCN)C(=O)N1CCC(N)(CC1)C(O)=O)NC(=O)[C@H](N)CC=1C=CC=CC=1)C1=CC=CC=C1 FWMNVWWHGCHHJJ-SKKKGAJSSA-N 0.000 description 6
- 108010026552 Proteome Proteins 0.000 description 6
- 238000002493 microarray Methods 0.000 description 6
- 238000007637 random forest analysis Methods 0.000 description 6
- 241000712461 unidentified influenza virus Species 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 5
- 210000004369 blood Anatomy 0.000 description 5
- 239000008280 blood Substances 0.000 description 5
- 201000004792 malaria Diseases 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- OBMZMSLWNNWEJA-XNCRXQDQSA-N C1=CC=2C(C[C@@H]3NC(=O)[C@@H](NC(=O)[C@H](NC(=O)N(CC#CCN(CCCC[C@H](NC(=O)[C@@H](CC4=CC=CC=C4)NC3=O)C(=O)N)CC=C)NC(=O)[C@@H](N)C)CC3=CNC4=C3C=CC=C4)C)=CNC=2C=C1 Chemical compound C1=CC=2C(C[C@@H]3NC(=O)[C@@H](NC(=O)[C@H](NC(=O)N(CC#CCN(CCCC[C@H](NC(=O)[C@@H](CC4=CC=CC=C4)NC3=O)C(=O)N)CC=C)NC(=O)[C@@H](N)C)CC3=CNC4=C3C=CC=C4)C)=CNC=2C=C1 OBMZMSLWNNWEJA-XNCRXQDQSA-N 0.000 description 4
- 208000001490 Dengue Diseases 0.000 description 4
- 206010012310 Dengue fever Diseases 0.000 description 4
- 101710176384 Peptide 1 Proteins 0.000 description 4
- 201000003486 coccidioidomycosis Diseases 0.000 description 4
- 208000025729 dengue disease Diseases 0.000 description 4
- UQLDLKMNUJERMK-UHFFFAOYSA-L di(octadecanoyloxy)lead Chemical compound [Pb+2].CCCCCCCCCCCCCCCCCC([O-])=O.CCCCCCCCCCCCCCCCCC([O-])=O UQLDLKMNUJERMK-UHFFFAOYSA-L 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 230000028993 immune response Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 239000000463 material Substances 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 235000018102 proteins Nutrition 0.000 description 4
- 102000004169 proteins and genes Human genes 0.000 description 4
- 239000000872 buffer Substances 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000001717 pathogenic effect Effects 0.000 description 3
- 238000000513 principal component analysis Methods 0.000 description 3
- 238000003908 quality control method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 201000008827 tuberculosis Diseases 0.000 description 3
- IGAZHQIYONOHQN-UHFFFAOYSA-N Alexa Fluor 555 Chemical compound C=12C=CC(=N)C(S(O)(=O)=O)=C2OC2=C(S(O)(=O)=O)C(N)=CC=C2C=1C1=CC=C(C(O)=O)C=C1C(O)=O IGAZHQIYONOHQN-UHFFFAOYSA-N 0.000 description 2
- 241000588832 Bordetella pertussis Species 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 2
- 238000000692 Student's t-test Methods 0.000 description 2
- 150000001413 amino acids Chemical class 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 239000000090 biomarker Substances 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000001413 cellular effect Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 239000012154 double-distilled water Substances 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 208000002672 hepatitis B Diseases 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 210000002381 plasma Anatomy 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 238000009589 serological test Methods 0.000 description 2
- 208000006379 syphilis Diseases 0.000 description 2
- 241000589968 Borrelia Species 0.000 description 1
- 208000017667 Chronic Disease Diseases 0.000 description 1
- 206010017533 Fungal infection Diseases 0.000 description 1
- 206010019799 Hepatitis viral Diseases 0.000 description 1
- 208000031888 Mycoses Diseases 0.000 description 1
- 229920001213 Polysorbate 20 Polymers 0.000 description 1
- 208000010362 Protozoan Infections Diseases 0.000 description 1
- 206010057190 Respiratory tract infections Diseases 0.000 description 1
- 206010040047 Sepsis Diseases 0.000 description 1
- 241001111950 Sonora Species 0.000 description 1
- YQNQNVDNTFHQSW-UHFFFAOYSA-N acetic acid [2-[[(5-nitro-2-thiazolyl)amino]-oxomethyl]phenyl] ester Chemical compound CC(=O)OC1=CC=CC=C1C(=O)NC1=NC=C([N+]([O-])=O)S1 YQNQNVDNTFHQSW-UHFFFAOYSA-N 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000000840 anti-viral effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000012148 binding buffer Substances 0.000 description 1
- 238000001574 biopsy Methods 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 210000001124 body fluid Anatomy 0.000 description 1
- 239000005018 casein Substances 0.000 description 1
- BECPQYXYKAMYBN-UHFFFAOYSA-N casein, tech. Chemical compound NCCCCC(C(O)=O)N=C(O)C(CC(O)=O)N=C(O)C(CCC(O)=N)N=C(O)C(CC(C)C)N=C(O)C(CCC(O)=O)N=C(O)C(CC(O)=O)N=C(O)C(CCC(O)=O)N=C(O)C(C(C)O)N=C(O)C(CCC(O)=N)N=C(O)C(CCC(O)=N)N=C(O)C(CCC(O)=N)N=C(O)C(CCC(O)=O)N=C(O)C(CCC(O)=O)N=C(O)C(COP(O)(O)=O)N=C(O)C(CCC(O)=N)N=C(O)C(N)CC1=CC=CC=C1 BECPQYXYKAMYBN-UHFFFAOYSA-N 0.000 description 1
- 235000021240 caseins Nutrition 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 210000001072 colon Anatomy 0.000 description 1
- 230000021615 conjugation Effects 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000007865 diluting Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 210000003608 fece Anatomy 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000002538 fungal effect Effects 0.000 description 1
- 238000011223 gene expression profiling Methods 0.000 description 1
- 238000011534 incubation Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003472 neutralizing effect Effects 0.000 description 1
- 229960002480 nitazoxanide Drugs 0.000 description 1
- 230000009871 nonspecific binding Effects 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 235000010486 polyoxyethylene sorbitan monolaurate Nutrition 0.000 description 1
- 239000000256 polyoxyethylene sorbitan monolaurate Substances 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 208000020029 respiratory tract infectious disease Diseases 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 210000003296 saliva Anatomy 0.000 description 1
- 239000012723 sample buffer Substances 0.000 description 1
- 238000013515 script Methods 0.000 description 1
- 210000000813 small intestine Anatomy 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 238000012353 t test Methods 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- UEUXEKPTXMALOB-UHFFFAOYSA-J tetrasodium;2-[2-[bis(carboxylatomethyl)amino]ethyl-(carboxylatomethyl)amino]acetate Chemical compound [Na+].[Na+].[Na+].[Na+].[O-]C(=O)CN(CC([O-])=O)CCN(CC([O-])=O)CC([O-])=O UEUXEKPTXMALOB-UHFFFAOYSA-J 0.000 description 1
- 210000002700 urine Anatomy 0.000 description 1
- 230000035899 viability Effects 0.000 description 1
- 244000052613 viral pathogen Species 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
- G01N33/50—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing
- G01N33/53—Immunoassay; Biospecific binding assay; Materials therefor
- G01N33/569—Immunoassay; Biospecific binding assay; Materials therefor for microorganisms, e.g. protozoa, bacteria, viruses
- G01N33/56911—Bacteria
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
- G01N33/50—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing
- G01N33/53—Immunoassay; Biospecific binding assay; Materials therefor
- G01N33/569—Immunoassay; Biospecific binding assay; Materials therefor for microorganisms, e.g. protozoa, bacteria, viruses
- G01N33/56983—Viruses
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2469/00—Immunoassays for the detection of microorganisms
- G01N2469/20—Detection of antibodies in sample from host which are directed against antigens from microorganisms
Definitions
- a method to distinguish a bacterial infection from a viral infection contains the steps of contacting an antibody-containing sample with an array of immobilized peptides, wherein said peptides are selected from a group consisting of one or more peptides that bind to antibodies produced in response to a bacterial infection and one or more peptides that bind to antibodies produced in response to a viral infection; and detecting binding of an antibody from said sample with a peptide from said group.
- FIG. 2 Panels A-D: Performance of distinguishing bacterial versus viral infection.
- Panel A PCA analysis on the selected peptides shows one factor is responsible for most variability, test set samples are highlighted in the right figure.
- Panel B Clustering of the selected peptides shows most peptides are bacteria specific peptides.
- Panel C Performance of the classification algorithms.
- Panel D Two selected peptides can achieve similar performance of classification.
- FIG. 5 Scatterplot of the 2 selected peptides. Color is true class. All samples are included in this figure. Both peptides are bacteria specific peptides.
- FIG. 6 Hierarchical clustering for the correlation of the whole immunosignature of each sample within bacterial and viral infections. More virus samples are misclassified as bacteria and mostly are influenza samples. Specificity for virus is nearly 100%.
- FIG. 7 Hierarchical clustering for the correlation of the whole immunosignature of each sample within bacterial and viral infections. More virus samples are misclassified as bacteria and mostly are influenza samples. Specificity for virus is nearly 100%.
- FIG. 8 Probability graph for being virus using Neural Network method in bacteria vs viral infection experiment. Color is true label. All samples are included in this figure. Graph shows good separation between the two groups.
- Immunosignature is a peptide microarray that derives peptide sequences from random space rather than biological sequence space.
- the analysis of semi-random sequences allows for a mostly unbiased search for antibodies that may display a common binding motif.
- the applicant would not focus on sequences for any given pathogen, which allows exploring more broadly for antibodies that may fall into a pattern that overlaps bacteria and virus.
- the immunosignatures of 40 different types of pathogens were examined. Each type of infection has a signature that distinguishes it from people without the infection and from other types of infection. Immunosignature by measuring the antibody response against pathogens, can distinguish bacterial from viral infections. Further, the applicant was able to identify 2 peptides that can distinguish the two classes, which would yield a biomarker with more clinical utility. Further, immunosignature can distinguish bacterial from generally a non-bacterial infection, which also has clinical relevance, since there are always non-bacterial and non-viral infections present in clinical settings.
- Immunosignatures are patterns of antibody binding on 125,000 peptide feature chips. The peptides are chosen from random peptide sequence space to maximize chemical diversity for discriminating antibody binding. Immunosignatures have been demonstrated to readily distinguish different types of infections and chronic diseases.
- a training set of x samples of bacteria, viral and non-infected were used to establish the signature. The training set was validated on an independent set of samples and then tested on another completely independent set. A set of 1000 peptides were identified that could make the distinction with X specificity and Y sensitivity in the test set. Most of the misclassifications were influenza samples called as bacterial infections.
- sample includes any biological specimen obtained from an individual. Suitable samples for use in the present invention include, without limitation, whole blood, plasma, serum, saliva, urine, stool (i.e., feces), tears, and any other bodily fluid, or a tissue sample (i.e., biopsy) such as a small intestine or colon sample, and cellular extracts thereof (e.g., red blood cellular extract).
- tissue sample i.e., biopsy
- the sample is a blood, plasma, or serum sample.
- sample is a serum sample. They represent a wide range of bacterial and virus species. There were between 9-22 sera samples from each type of pathogen.
- FIG. 1 demonstrates the initial unsupervised division showing that influenza virus is the sole misclassified group, classified with bacteria.
- FIG. 6 A further breakdown per samples is shown in FIG. 6 .
- Hierarchical clustering using the correlations for every sample (no sample is averaged) is shown in FIG. 6 .
- the specificity for viral infections is close to 100%, with some viruses being classified as bacteria, mostly influenza. This result is consistent with the class level clustering result.
- Example 1 Building Bacterial Versus Viral Infection Classifier that Shows Robust Distinction
- the diagnostic peptides are chosen against a background of non-diagnostic peptides on the arrays.
- the diagnostic peptides can be arrayed with a set of 100-10,000 random peptides. These could be mixed or individual and spotted separately or as a mixture.
- PCA Principle Component Analysis
- test set samples are also highlighted in the clustering heatmap to show their clustering group location compared with the training and validation set. No obvious overfitting is noted as test set samples are generally clustered in the right class.
- Machine learning classifiers like Random Forest and Neural Networks are used to build the model of classification between the two groups.
- a model is trained using training data and a validation set is used to fine-tune the model and gain an initial performance evaluation to limit overfitting.
- training group only usually results in overfitting because the classifier might adjust to the random variations in the training group to gain best fit scores.
- Validation set only also pose the same issue because the model is generated with information from the validation dataset. In microarray studies, there are inevitably more variables than observations, overfitting becomes more pronounced. Independent datasets are needed to test the performance of the classifier, the test set data are never used in feature selection to model generating and is only used for the final evaluation of the model.
- Random Forest tends to exhibit less sensitivity to the bacterial infections (sensitivity at 0.58) but is extremely specific (0.95). This is a bias toward true negatives as the cost of lower true positives.
- Neural Network models yielded more balance for TP and FP between the two groups, with sensitivity and specificity at 0.83 and 0.84 respectively ( FIG. 8 ). Both models yield misclassification rates of less than 20%. Since it may be that up to 60% of human infections are from virus, if doctors can distinguish viral infections from bacterial infections, the use of antibiotics could be reduced by over 50%.
- Stepwise regression is utilized to find the optimal, non-redundant peptides that can be used to fit the model.
- Each peptide has to meet a p-value cutoff of 0.1 to enter the model and will exit the model upon the exceeding the existing cutoff p-value of 0.1.
- Regression is started assuming all peptides are out of the model. The whole process is iterated until the model is stabilized, meaning, i.e., no peptides leave or enter the model. Then the model is fine-tuned to maximize RSquare for the Validation set ( FIG. 2 , panel D).
- the final model only includes two peptides, GALSRSFANVSFPGVAG (SEQ ID NO: 1) and GLSNGASSFGKASGVAL (SEQ ID NO: 2) ( FIG. 5 ). Specificity and sensitivity for the test set comes to 0.75 and 0.89, only marginally worse than the complete models using all 1000 peptides. And the misclassification rate is at 0.16, no worse than the complete models.
- Example 2 Epitopes of Bacteria were Identified Via Blast Search of the Two Peptides followeded by Ungapped Motif Mapping
- the two peptides may contain bacterial epitopes or mimotopes that enhance bacteria-specific antibody binding.
- One hundred matched sequences in bacteria proteomes were identified and subsequently submitted to the MEME tool in the MEME suite to identify consensus motifs.
- the identified motif(s) are the epitope(s) from bacteria that the 2 peptides represent. Results are shown in table 3.
- One epitope (SEQ ID NO: 3) is identified for peptide 1 (SEQ ID NO: 1); and two epitopes (SEQ ID NO: 4 and SEQ ID NO: 5) were identified for peptide 2 (SEQ ID NO: 2). It is interesting to note that for peptide 1 (SEQ ID NO: 1), only 6 amino acids seem to be the target of bacterial specific antibodies. As for peptide 2 (SEQ ID NO: 2), the full length of the peptide could be the target of bacterial specific antibodies. Each epitope is matched with at least 20 sequences from the bacterial proteome, so the epitopes are broadly represented in the bacterial world.
- Peptide 1 (SEQ ID NO: 1) has 1 epitope (SEQ ID NO: 3) with length of 6 amino acid (a.a). While peptide 2 (SEQ ID NO: 2) has 2 matched epitopes (SEQ ID No. 4 and SEQ ID NO: 5) with length of 8 a.a. and 6 a.a. correspondingly. Matched part is highlighted with color in peptides. This implies only part of peptide 1 is identified by bacterial specific antibody while the whole sequence of peptide 2 is the target for bacterial antibodies.
- epitope 1 epitope 2 GALSRSFANVSFPGVAG RSFANV (SEQ ID NO: 1) (SEQ ID NO: 3) GLSNGASSFGKASGVAL SFGKASGV LSNGAS (SEQ ID NO: 2) (SEQ ID NO: 4) (SEQ ID NO: 5)
- Example 3 Broad Bacterial Versus Non-Bacterial Infection Classifier Shows Robust Distinction and Better Performance
- the better performing Neural Network model is at 0.83 sensitivity and 0.94 specificity for bacteria with a Generalized RSquare of 0.73, all of which is a significant increase compared to the original bacterial vs viral model.
- This improvement can be the result of more samples being used for model construction, or it can be by including more types of infection as the non-bacteria comparison, which can lead to a more robust bacterial specific signature.
- Immunosignature a microarray-based serological test that uses semi-random peptides to splay out the antibody repertoire from infected individuals, is used to distinguish viral infections from bacterial infections.
- Immunosignatures can detect peptides that generally separate bacterial infections from viral infections.
- Machine learning models were used to identify the predictive performance of a given set of peptides across 169 patients, of which 105 patients have bacterial infections and 64 viral infections. We achieved over 84% accuracy, 84% specificity, and 83% sensitivity, and could achieve similar accuracy, specificity, and sensitivity with as few as two peptides. These two peptides were overrepresented in bacterial proteomes, and underrepresented in viral proteomes. Even when adding fungal and protozoan infections, high specificity is maintained, an important goal to achieve when attempting to reduce improperly prescribed antibiotics.
- Correlation of the infections are used to first test the possibility of distinction at the antibody system level.
- the logic behind using correlation of infections is that the immune system might systematically see the difference between bacterial and viral infection by activating different pathways.
- Immunosignature platforms measure antibody repertoire in the blood. If you use all the data from the platform, then you are measuring the immune system. Correlation of the immune system can then be tested by calculating the correlation of the immunosignature for different pathogens.
- the result from the correlation offer insights into understanding both diagnosis and how the immune system works. It seems the immune system is able to distinguish most bacterial and viral infections and mount totally different immune response, since only one infection is misclassified. This confirms the notion that our immune system probably knows the source of the infection and respond accordingly.
- the immune system does not know the source of infection but because all infections within the same class are so similar, the immune system always produces similar antibodies against various bacterial infections. The same might be the case for viral infection.
- most of the signatures that can distinguish bacterial and viral infection are bacterial specific signatures, implying the immune system is producing various antibodies against bacterial infection in ways analogous to broad-spectrum antibiotics.
- influenza virus is misclassified into bacteria is interesting because it suggests somehow influenza virus successfully tricked the immune system into thinking it is bacteria such that antibodies against bacteria are produced, the result of which will be ineffective. This is consistent with the fact that viruses are highly contagious worldwide, implying the immune system cannot quickly mount an effective immune response because influenza virus is regarded as “bacteria.” This misclassification by the immune system might also explain why there are already pre-existing neutralizing antibodies within the immune system, but they were not usually elucidated during flu infection.
- Serum samples were collected at various sources described in detail below and received at Arizona State University (ASU) under Institutional Review Board Protocol #0912004625, “Profiling Serum for Unique Antibody Signatures”. All samples have informed consent and were anonymized. Every disease sample was tested positive for the specified disease before rendering to ASU. Bordetella pertussis samples were provided by Seracare Life Sciences (Seracare). Tuberculosis from University of Texas at El Paso (UTEP). Malaria from Seracare. HIV from Creative Testing Solutions (CTS). Flu from BioreclamationlVT. Dengue from UTEP. WNV from CTS. VF from Sonora Lab. Chagas from CTS. Lyme from Seracare. Hepatitis B from CTS, Syphilis from Seracare.
- Bordetella pertussis , Lyme, Syphillis, Tuberculosis, Dengue, Flu, Hepatitis B, HIV and WNV samples are used in the bacterial versus viral experiment. Chagas, Malaria and Valley Fever were added in the bacterial versus non-bacterial experiment. All samples are randomly assigned into training, validation and test set with equal probability.
- Serum samples were diluted 1:1500 into the sample buffer (3% BSA in 1 ⁇ PBST) before incubated on immunosignature microarrays at a final volume of 150 ul for 1 h at 37° C. with rotating. Primary antibodies from the serum were then washed with 1 ⁇ PBST for 3 times and rinsed with ddH 2 O for 3 times. 4 nM Secondary anti-human IgG antibodies with Alexa-Fluor 555 conjugation from Life Technologies are added in secondary incubation buffer (0.75% Casein in 1 ⁇ PBST with 0.05% Tween20) to detect primary antibody binding. Secondary antibodies were incubated on the array for 1 h at 37° C. before washed off with blocking buffer. Slides were then washed with 1 ⁇ PBST and ddH 2 O before drying. Images were obtained from scanning arrays at 555 nm using Innoscan 910 scanner. Signal intensity for features were extracted using GenePix Pro 6.0.
- Feature selection is done by using samples in the training and validation set. Two-tail Student's T-Test is performed for each peptides by comparing bacterial infection samples versus viral infection samples (non-bacterial infection samples). Cutoff is controlled at allowing 1 false positive for all test, which is 1/124,000 or 1000 peptides, whichever is smaller.
- PCA is performed using selected peptides with all samples, with the test set samples highlighted in right PCA plot.
- Hierarchical clustering is performed using the selected peptides with all samples. Ward method is used in calculating the distance between the samples. The same method is used in calculating distance for the features in two-way clustering.
- Random Forest is carried out with maximum 100 trees in the forest. Minimum split per tree is set at 10 and maximum at 2000. Early stopping rule is applied on validation set. And performance of the classifier is evaluated and output as confusion matrix for the training, validation and test set.
- Neural Network is built with one hidden layer and 3 nodes, with Tan H as the activation function.
- Stepwise regression for reducing number of features is used with stopping rule of p-value cutoff at 0.1 for both entering and leaving the model.
- the model starts empty with no feature.
- Features become included in the model if below cutoff p-value and will be removed from the model once p-value larger than the cutoff. This process is done recursively until the model stabilize, with no feature entering and leaving the model. Then the selected features are tuned to maximum RSquare for the validation set. Then Logistic regression is used in building model with the 2 selected peptides.
- Blast search of the 2 peptides is done using the NCBI blast server.
- Protein Blast (blastp) suite is used.
- Database is Reference proteins and organism is limited to Bacteria (taxid:2).
- Algorithm parameters is set to adjust for short sequences, and max target sequences at 100. Then the matched sequences are processed to contain only linear matched part.
- the 100 matched sequences are imported into MEME suite to identify epitopes, with configurations of 10 minimum sites per epitope and 3 maximum epitopes.
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Immunology (AREA)
- Engineering & Computer Science (AREA)
- Urology & Nephrology (AREA)
- Chemical & Material Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Hematology (AREA)
- Virology (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Tropical Medicine & Parasitology (AREA)
- Microbiology (AREA)
- Food Science & Technology (AREA)
- Cell Biology (AREA)
- Physics & Mathematics (AREA)
- Biotechnology (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Pathology (AREA)
- Peptides Or Proteins (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Abstract
Assays, arrays, and methods for distinguishing a bacterial infection from a viral infection are disclosed. The antibiotic crisis is in part driven by over prescription of antibiotics. There is a tendency, particular in pediatrics, to give an antibiotic even for viral infections. Thus, embodiments herein are directed to the problem of distinguishing a bacterial infection from a viral infection to reduce unnecessary antibiotic usage.
Description
- This application is a continuation of U.S. application Ser. No. 16/896,117, filed Jun. 8, 2020, which is a continuation of U.S. application Ser. No. 16/480,887, filed Jul. 25, 2019, and issued on Jul. 14, 2020 as U.S. Pat. No. 10,712,342, which represents the national stage entry of PCT International Application No. PCT/US2018/016185, filed on Jan. 31, 2018, and is based on, and claims a priority benefit from U.S. Provisional Patent Application No. 62/452,825, filed Jan. 31, 2017, and entitled “Diagnostic to Distinguish Bacterial Infections,” each of which is incorporated herein by reference in its entirety.
- This invention was made with government support under HSHQDC-15-C-B0008 awarded by the Department of Homeland Security, Science and Technology. The government has certain rights in the invention.
- The content of the ASCII text file of the sequence listing named “Sequence_Listing_CALV_036C2.txt” which is 1.38 kb in size was created on May 31, 2022 and electronically submitted via EFS-Web herewith the application is incorporated herein by reference in its entirety.
- Antibiotic resistance is a global problem mainly due to the overuse of antibiotics in clinical settings. Overuse is mainly caused by the lack of accurate diagnosis that can distinguish bacterial infections from other types of infections. This is especially true for respiratory tract infections and pediatric sepsis. More accurate diagnosis at the time of an initial clinical visit that can distinguish bacterial from other infections would greatly curb the antibiotic overuse problem.
- A major advance in stemming the antibiotic crisis would be to have a diagnostic that could readily distinguish a bacterial from viral infection on presentation with symptoms. This would decrease the unnecessary use of antibiotics while still allowing their application optimally for bacterial infections. Current research on distinguishing bacterial from viral infections has mostly been focusing on genome-wide expressions (GWAS). The notion is that gene expression will change upon infections of different pathogens. However, a serological test detection method for pathogens is antibody response. There are many complicating factors that make analysis of antibodies between viral and bacterial infections complex—one of the most important is the study platform.
- Embodiments of the current disclosure describe an array and methods for distinguishing a bacterial infection from a viral infection. In certain embodiments, the array comprises two peptides, which first peptide comprises SEQ ID NO: 1 and said second peptide comprises SEQ ID NO: 2. Further, the first peptide comprises a motif able to be bound to a plurality of bacterial specific antibodies, wherein the motif comprises SEQ ID NO: 3 and the second peptide comprises a first motif and a second motif able to be bound to a plurality of bacterial specific antibodies, wherein the first motif comprises SEQ ID NO: 4 and the second motif comprises SEQ ID NO: 5.
- In certain embodiments, a method to distinguish a bacterial infection from a viral infection is disclosed. The method contains the steps of contacting an antibody-containing sample with an array of immobilized peptides, wherein said peptides are selected from a group consisting of one or more peptides that bind to antibodies produced in response to a bacterial infection and one or more peptides that bind to antibodies produced in response to a viral infection; and detecting binding of an antibody from said sample with a peptide from said group.
- Embodiments of the invention will be better understood from a reading of the following detailed description taken in conjunction with the drawings in which like reference designators are used to designate like elements, and in which:
-
FIG. 1 . Hierarchical clustering for the correlation of the whole immunosignature by type of infection shows potential classification of bacterial versus viral infection. Correlation is calculated for each pair-wise sample comparison, then the samples that belong to the same class are averaged to a single correlation value. The clustering table shows most viruses can be distinguished from the bacteria, with the exception of flu. -
FIG. 2 , Panels A-D: Performance of distinguishing bacterial versus viral infection. Panel A: PCA analysis on the selected peptides shows one factor is responsible for most variability, test set samples are highlighted in the right figure. Panel B: Clustering of the selected peptides shows most peptides are bacteria specific peptides. Panel C: Performance of the classification algorithms. Panel D: Two selected peptides can achieve similar performance of classification. -
FIG. 3 , Panels A-C: Performance of distinguishing bacterial versus viral and other types of infection. Panel A: PCA analysis on the selected peptides shows one factor is responsible for most variability, test set samples are highlighted in the right figure. Panel B: Clustering of the selected peptides shows most peptides are bacteria specific peptides. Panel C: Performance of the classification algorithms. -
FIG. 4 . Hierarchical clustering for the correlation of the whole immunosignature by type of infection including all classes. Non-infected class is more similar to bacterial infection, while the non-bacterial and non-viral infections are spread out in groups. -
FIG. 5 . Scatterplot of the 2 selected peptides. Color is true class. All samples are included in this figure. Both peptides are bacteria specific peptides. -
FIG. 6 . Hierarchical clustering for the correlation of the whole immunosignature of each sample within bacterial and viral infections. More virus samples are misclassified as bacteria and mostly are influenza samples. Specificity for virus is nearly 100%. -
FIG. 7 . Hierarchical clustering for the correlation of the whole immunosignature of each sample within bacterial and viral infections. More virus samples are misclassified as bacteria and mostly are influenza samples. Specificity for virus is nearly 100%. -
FIG. 8 . Probability graph for being virus using Neural Network method in bacteria vs viral infection experiment. Color is true label. All samples are included in this figure. Graph shows good separation between the two groups. - Embodiments of the disclosure are described in preferred embodiments in the following description with reference to the Figures, in which like numbers represent the same or similar elements. Reference throughout this specification to “one embodiment,” “an embodiment,” or similar language means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, appearances of the phrases “in one embodiment,” “in an embodiment,” and similar language throughout this specification may, but do not necessarily, all refer to the same embodiment.
- The described features, structures, or characteristics of the invention may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are recited to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention may be practiced without one or more of the specific details, or with other methods, components, materials, and so forth. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the invention.
- Immunosignature is a peptide microarray that derives peptide sequences from random space rather than biological sequence space. The analysis of semi-random sequences allows for a mostly unbiased search for antibodies that may display a common binding motif. The applicant would not focus on sequences for any given pathogen, which allows exploring more broadly for antibodies that may fall into a pattern that overlaps bacteria and virus.
- The immunosignatures of 40 different types of pathogens were examined. Each type of infection has a signature that distinguishes it from people without the infection and from other types of infection. Immunosignature by measuring the antibody response against pathogens, can distinguish bacterial from viral infections. Further, the applicant was able to identify 2 peptides that can distinguish the two classes, which would yield a biomarker with more clinical utility. Further, immunosignature can distinguish bacterial from generally a non-bacterial infection, which also has clinical relevance, since there are always non-bacterial and non-viral infections present in clinical settings.
- Pooled materials were examined for common immunosignatures between 4 different bacteria and 5 different viruses, in total 280 samples. Immunosignatures (IMS) are patterns of antibody binding on 125,000 peptide feature chips. The peptides are chosen from random peptide sequence space to maximize chemical diversity for discriminating antibody binding. Immunosignatures have been demonstrated to readily distinguish different types of infections and chronic diseases. A training set of x samples of bacteria, viral and non-infected were used to establish the signature. The training set was validated on an independent set of samples and then tested on another completely independent set. A set of 1000 peptides were identified that could make the distinction with X specificity and Y sensitivity in the test set. Most of the misclassifications were influenza samples called as bacterial infections. To further explore the limits of IMS we included samples from 3 eukaryotic pathogens. While the ability to distinguish all 3 classes decreased in accuracy, the accuracy of distinguishing bacterial from virus and eukaryotic pathogen increased. To examine the issue of whether a lesser number of peptides could distinguish bacterial from viral infection, each peptide was tested for performance. We determined that two peptides performed as well as the 1000 in making the bacteria versus virus call. These two peptides contained motifs that were common in bacterial proteomes. We used the natural bacterial peptides in a simple spot assay to test bacterial and viral sera samples and demonstrated preliminary distinction of the samples.
- The sera samples used are listed in Table 1. The term “sample” includes any biological specimen obtained from an individual. Suitable samples for use in the present invention include, without limitation, whole blood, plasma, serum, saliva, urine, stool (i.e., feces), tears, and any other bodily fluid, or a tissue sample (i.e., biopsy) such as a small intestine or colon sample, and cellular extracts thereof (e.g., red blood cellular extract). In a preferred embodiment, the sample is a blood, plasma, or serum sample. In a more preferred embodiment, the sample is a serum sample. They represent a wide range of bacterial and virus species. There were between 9-22 sera samples from each type of pathogen. Each sample was run in duplicate on the standard CIMV7 arrays containing 125K peptides. The process has been described, but briefly, involves diluting the sample 600× in buffer, applying it to the array for one hour, washing and then detecting the pattern of antibody binding with a labeled secondary. In the assays reported here, IgG was detected.
-
TABLE 1 Sample information used in this application. 12 classes of infections are included in addition to a group of non-infected individuals coded as normal. Class of Type of Sample Count per infection infection number class non-infected non-infected 62 62 Bacteria Borrelia 9 64 Bacteria Lyme 13 Bacteria Syphilis 22 Bacteria Tuberculosis 20 Virus Dengue 22 105 Virus Flu 22 Virus Hepatitis B 20 Virus HIV 21 Virus WNV 20 Other Chagas 19 57 Other Malaria 17 Other Valley Fever 21 - If the immune system responds to bacterial and viral infections differently, a high correlation for the immune responses within each group and low correlation between them is expected. Correlation is calculated for each pair-wise sample comparison with the 125 k features on the immunosignature, then the samples belong to the same class are averaged to a single correlation value (
FIG. 1 ). For example, correlations for all comparisons between any Dengue samples versus any WNV samples are averaged into a single value, representing the average correlation between the two groups. Hierarchical clustering results show bacteria and virus are separated, with the only exception of influenza virus, which is classified with bacteria.FIG. 1 demonstrates the initial unsupervised division showing that influenza virus is the sole misclassified group, classified with bacteria. Given the fact that there is a lot of noise in the immune system and numerous irrelevant antibodies circulating in the blood, this performance exceeds expectation and confirms the hypothesis that the immune system is able to distinguish bacterial and viral infections by producing different antibodies. Non-infected samples and non-bacterial non-viral pathogens are mixed when including them in the correlation table (FIG. 4 ). - A further breakdown per samples is shown in
FIG. 6 . Hierarchical clustering using the correlations for every sample (no sample is averaged) is shown inFIG. 6 . The specificity for viral infections is close to 100%, with some viruses being classified as bacteria, mostly influenza. This result is consistent with the class level clustering result. - Since samples in one class are merged, performance of each sample is determined. Hierarchical clustering for the correlation table for every sample is shown in
FIG. 3 panel A. And the specificity for virus is close to 100%, with some viruses being classified as bacteria, mostly influenza viruses. This result is consistent with the per-class clustering result. - Once the viability of distinguishing the two types of infections was confirmed, we utilized machine learning techniques to classify the samples. In this experiment, only bacterial and viral infection samples are used, with a total of 157 samples. Experimental workflow is outlined in
FIG. 7 . All samples are randomly divided into training, validation and held-out test set, with a ratio of 60%, 20%, 20%. Training and validation sets are used to build the classifier. Test sets remains untouched until the final model is constructed and used only for evaluation. - Since there are 125,000 features on immunosignature platform, it is plausible to first do feature selection to find the most useful peptides and remove noise. Feature selection is performed using training and validation set data via two-tail t-test for every peptide and top 1000 significant peptides are used. Note that the general cutoff is either selecting top 1000 peptides or p-value<1/125,000, controlling overall false positive sample to be less than 1. Whichever cutoff has smaller peptide numbers is used in the experiment. For tests we performed, the p-values are much lower than 1/125,000. As a result, a common cutoff of the top 1000 peptides is used throughout the application. In certain embodiments, the diagnostic peptides are chosen against a background of non-diagnostic peptides on the arrays. In order to provide the non-specific binding buffer to increase specificity, the diagnostic peptides can be arrayed with a set of 100-10,000 random peptides. These could be mixed or individual and spotted separately or as a mixture.
- Using the selected features, Principle Component Analysis (PCA) is performed to determine how many components are responsible for the majority of the variability (
FIG. 2 , panel A). Interestingly, it was found thatcomponent 1 alone explains over 60% of the variability, indicating at least one factor is strongly driving the variance across groups, at least for the selected features. The test samples are not used in feature selection, however, when analyzed with PCA (highlighted inFIG. 2 , panel A) the test set samples are well separated at the validation set would suggest, suggesting overfitting is negligible. Hierarchical clustering is performed using the selected features to visualize the data (FIG. 2 , panel B). As we can see most peptides are relatively higher in intensity in bacterial than viral infections. This suggests the peptides that are being selected are from antibody response were raised to the bacterial infection. The test set samples are also highlighted in the clustering heatmap to show their clustering group location compared with the training and validation set. No obvious overfitting is noted as test set samples are generally clustered in the right class. - Machine learning classifiers like Random Forest and Neural Networks are used to build the model of classification between the two groups. For each classifier, a model is trained using training data and a validation set is used to fine-tune the model and gain an initial performance evaluation to limit overfitting. After the established model is used on the test set, we perform a final performance evaluation on this independent dataset. Experiments with training group only usually results in overfitting because the classifier might adjust to the random variations in the training group to gain best fit scores. Validation set only also pose the same issue because the model is generated with information from the validation dataset. In microarray studies, there are inevitably more variables than observations, overfitting becomes more pronounced. Independent datasets are needed to test the performance of the classifier, the test set data are never used in feature selection to model generating and is only used for the final evaluation of the model.
- As it is shown in
FIG. 2 , panel C, Random Forest and Neural Networks both have minimal misclassification rate on both training and validation. The final performance on the test set is also similar for both classifiers. Random Forest tends to exhibit less sensitivity to the bacterial infections (sensitivity at 0.58) but is extremely specific (0.95). This is a bias toward true negatives as the cost of lower true positives. Neural Network models yielded more balance for TP and FP between the two groups, with sensitivity and specificity at 0.83 and 0.84 respectively (FIG. 8 ). Both models yield misclassification rates of less than 20%. Since it may be that up to 60% of human infections are from virus, if doctors can distinguish viral infections from bacterial infections, the use of antibiotics could be reduced by over 50%. - Stepwise regression is utilized to find the optimal, non-redundant peptides that can be used to fit the model. Each peptide has to meet a p-value cutoff of 0.1 to enter the model and will exit the model upon the exceeding the existing cutoff p-value of 0.1. Regression is started assuming all peptides are out of the model. The whole process is iterated until the model is stabilized, meaning, i.e., no peptides leave or enter the model. Then the model is fine-tuned to maximize RSquare for the Validation set (
FIG. 2 , panel D). The final model only includes two peptides, GALSRSFANVSFPGVAG (SEQ ID NO: 1) and GLSNGASSFGKASGVAL (SEQ ID NO: 2) (FIG. 5 ). Specificity and sensitivity for the test set comes to 0.75 and 0.89, only marginally worse than the complete models using all 1000 peptides. And the misclassification rate is at 0.16, no worse than the complete models. - Upon performing blast search on these 2 peptides against the RefSeq database excluding Homo sapiens, Models (XM/XP) and Uncultured/environmental sample sequences, they were found to be highly enriched in bacteria but not in virus (
FIG. 4 ). Furthermore, they are prevalent in all types of bacteria and all types of proteins, suggesting they are indeed good bacterial infection detection molecules/biomarkers. -
TABLE 2 Performance of bacterial vs non-bacterial infection classification using 5 selected peptides. Peptides are selected from stepwise regression using mixed p-value model at cutoff of 0.1. Logistic fit is then performed using the selected peptides. Test set performance is much lower compared with the complete model using all selected peptides from T-Test. logistic Fit Training Validation Test Sample size 127 42 43 Misclassification rate 0.06 0.14 0.23 sensitivity(Bacteria) 0.89 0.58 0.45 Specificity(Bacteria) 0.96 0.97 0.875 - Epitopes within the two peptides were examined. The two peptides may contain bacterial epitopes or mimotopes that enhance bacteria-specific antibody binding. We then did a protein blastp search of the 2 peptides against the Bacteria (taxid:2) with no E-value cutoff. One hundred matched sequences in bacteria proteomes were identified and subsequently submitted to the MEME tool in the MEME suite to identify consensus motifs. The identified motif(s) are the epitope(s) from bacteria that the 2 peptides represent. Results are shown in table 3. One epitope (SEQ ID NO: 3) is identified for peptide 1 (SEQ ID NO: 1); and two epitopes (SEQ ID NO: 4 and SEQ ID NO: 5) were identified for peptide 2 (SEQ ID NO: 2). It is interesting to note that for peptide 1 (SEQ ID NO: 1), only 6 amino acids seem to be the target of bacterial specific antibodies. As for peptide 2 (SEQ ID NO: 2), the full length of the peptide could be the target of bacterial specific antibodies. Each epitope is matched with at least 20 sequences from the bacterial proteome, so the epitopes are broadly represented in the bacterial world.
-
TABLE 3 Identified epitopes of bacteria with the 2 bacterial-viral distinguishing peptides. Peptide 1 (SEQ ID NO: 1) has 1 epitope (SEQ ID NO: 3) with length of 6 amino acid (a.a). While peptide 2 (SEQ ID NO: 2) has 2 matched epitopes (SEQ ID No. 4 and SEQ ID NO: 5) with length of 8 a.a. and 6 a.a. correspondingly. Matched part is highlighted with color in peptides. This implies only part of peptide 1 is identified by bacterial specificantibody while the whole sequence of peptide 2 is the targetfor bacterial antibodies. epitope 1epitope 2GALSRSFANVSFPGVAG RSFANV (SEQ ID NO: 1) (SEQ ID NO: 3) GLSNGASSFGKASGVAL SFGKASGV LSNGAS (SEQ ID NO: 2) (SEQ ID NO: 4) (SEQ ID NO: 5) - In clinical settings, it is possible that one will encounter nonbacterial or non-viral infections, so the ability to distinguish the other types of infections is important in doing a correct diagnosis. Thus, the focus is on binary classification of bacterial vs non-bacterial infections, because patients with bacterial infection can immediately receive antibiotics as treatment while other infections need more detail to arrive at a disease specific treatment. Accordingly, for example, Chagas, malaria and Valley Fever are included as noise and combined with viral infections as the non-bacterial infection class.
- Experiments were performed as described in
FIG. 7 , samples are divided into training, validation and test set. Training and validation set are used to do feature selection and construct a model, then performance is tested on the independent test set. Results were summarized inFIG. 3 , panels A-C. PCA analysis (FIG. 3 , panel A) and hierarchical clustering (FIG. 3 , panel B) show similar separation of the two group likeFIG. 2 , panels A-D, suggesting performance does not deteriorate when noise is added. Random Forest model and Neural Network model revealed a misclassification rate of 0.12 and 0.09 for the test set, which is an improvement compared with the bacterial vs viral only model. The better performing Neural Network model is at 0.83 sensitivity and 0.94 specificity for bacteria with a Generalized RSquare of 0.73, all of which is a significant increase compared to the original bacterial vs viral model. This improvement can be the result of more samples being used for model construction, or it can be by including more types of infection as the non-bacteria comparison, which can lead to a more robust bacterial specific signature. - In this experiment, we also attempted to find minimal number of peptides that can achieve similar performance compared with using all selected peptides. However, after the same stepwise regression process, the best performance we can get is using 5 peptides to gain a misclassification rate of 0.23, which is not as good as the complete model using all 1000 peptides. Also, the sensitivity for bacteria only was 0.44, also significantly lower than the Neural Network model.
- Immunosignature, a microarray-based serological test that uses semi-random peptides to splay out the antibody repertoire from infected individuals, is used to distinguish viral infections from bacterial infections. Immunosignatures can detect peptides that generally separate bacterial infections from viral infections. Machine learning models were used to identify the predictive performance of a given set of peptides across 169 patients, of which 105 patients have bacterial infections and 64 viral infections. We achieved over 84% accuracy, 84% specificity, and 83% sensitivity, and could achieve similar accuracy, specificity, and sensitivity with as few as two peptides. These two peptides were overrepresented in bacterial proteomes, and underrepresented in viral proteomes. Even when adding fungal and protozoan infections, high specificity is maintained, an important goal to achieve when attempting to reduce improperly prescribed antibiotics.
- Accurate diagnosis of bacterial and viral infections is needed in clinical settings. The current imprecise diagnosis results in either over use of antibiotics or delayed treatment for patients. Herein is a novel diagnosis based on immunosignature technology that is able to reliably diagnose bacterial infection from viral infections. By measuring the antibody response of patients with different infections, the ability to distinguish the majority of the bacterial and viral infections has been demonstrated. We further construct models based on selected features and applying machine learning algorithms to the selected features. This model is able to classify the two types of infections with misclassification rate of less than 20%, exceeding current methods used either in research or clinical settings. Since in clinical settings non-bacterial, non-viral infections will be expected, we also constructed a model aimed at distinguishing bacterial versus all other non-bacterial infections, consisting of viral infection and noise infections including Chagas, Malaria and Valley Fever. This model shows even better performance with misclassification rate at about 10%.
- Several studies using gene expression profiling have shown potential to diagnose of bacterial vs viral infections. The logic behind those studies is genes will be differentially regulated when encountering different infections. So is it the case for antibody response. Antibody response is the most direct reaction for an infection. Given the fact that genes as indirect reactions can still work to distinguish infections types, antibody response should be an even better approach because of it directly targeting the pathogens. One thing worth noting is that compared with gene microarrays, where it is usually one-to-one binding, antibodies will usually bind to multiple peptides on an immunosignature platform as long as the peptides are mimotopes of the true epitope. As a result, more peptides are used in analysis for the immunosignature experiments.
- Correlation of the infections are used to first test the possibility of distinction at the antibody system level. The logic behind using correlation of infections is that the immune system might systematically see the difference between bacterial and viral infection by activating different pathways. Immunosignature platforms measure antibody repertoire in the blood. If you use all the data from the platform, then you are measuring the immune system. Correlation of the immune system can then be tested by calculating the correlation of the immunosignature for different pathogens. The result from the correlation offer insights into understanding both diagnosis and how the immune system works. It seems the immune system is able to distinguish most bacterial and viral infections and mount totally different immune response, since only one infection is misclassified. This confirms the notion that our immune system probably knows the source of the infection and respond accordingly. Or perhaps the immune system does not know the source of infection but because all infections within the same class are so similar, the immune system always produces similar antibodies against various bacterial infections. The same might be the case for viral infection. As discussed above, most of the signatures that can distinguish bacterial and viral infection are bacterial specific signatures, implying the immune system is producing various antibodies against bacterial infection in ways analogous to broad-spectrum antibiotics.
- And the result that influenza virus is misclassified into bacteria is interesting because it suggests somehow influenza virus successfully tricked the immune system into thinking it is bacteria such that antibodies against bacteria are produced, the result of which will be ineffective. This is consistent with the fact that viruses are highly contagious worldwide, implying the immune system cannot quickly mount an effective immune response because influenza virus is regarded as “bacteria.” This misclassification by the immune system might also explain why there are already pre-existing neutralizing antibodies within the immune system, but they were not usually elucidated during flu infection.
- Overfitting has been a major problem in microarray studies. Here we approach the experiments with a pre-isolated test set data to avoid the problem. The whole model construction process is without information from the test set. After the model is stabilized, its performance is tested with the test set data. And the results show there is little overfitting when migrating the model from the training, validation set to the test set.
- In the bacterial versus viral infection model, we are achieving accuracy of over 80% in both classifiers tested. And clinicians can choose which classifier to use based on experience, since following the random forest classifier will minimize the diagnosis of viral infection into bacterial infection, hence lower the usage of anti-biotics, while the neural network classifier tends to balance the error rate in each class, resulting in more usage of antibiotics but less suffering of patients who genuinely have a bacterial infection. Features being selected from this study are almost exclusively from bacterial infection, indicating there is more commonality with the immune response.
- The two peptides were further examined by identifying matched sequences from bacteria proteomes and then identifying consensus motifs with the matched sequences. These consensus motifs could be the binding target within the two peptides on the immunosignature. Only 6 a.a. of consensus motif in one of the peptides is identified and the full length of the other peptide is matched by bacterial antibodies. This indicates there could be redundancy in these two peptides.
- Interestingly, when non-bacterial and non-viral infections are added as the non-bacterial class, the performance of the model actually increases. Accuracy is at −90% in both classifiers. And specificity for Bacteria it is −95% in both classifiers, indicating this model is good at distinguishing non-bacterial infections. When coupled with the results of the clustering heatmap, it appears that our immune system sees the commonality for bacterial infections but not other types. This is interpreted from the classifier result that all features are bacteria specific features and as long as you don't have those features, you are classified into the non-bacterial class.
- In summary, we are able to construct classifiers that are better performing for bacterial versus viral infection. We validated each model using independent dataset to confirm the robustness of the model. And we are able to confirm the source of the selected features, which in turn offers the logic for the success of the model. We believe immunosignatures can be beneficial when used in clinical settings to both combat the antibiotic overdose problem and to reduce suffering of the patients. In other words, a patient whose antibody-containing sample binds to one or more peptides indicating a bacterial infection can be treated with antibiotics, while a patient whose antibody-containing sample binds to one or more peptides on the array indicating a viral infection can be treated with an anti-viral medicament, such as nitazoxanide.
- Serum samples were collected at various sources described in detail below and received at Arizona State University (ASU) under Institutional Review Board Protocol #0912004625, “Profiling Serum for Unique Antibody Signatures”. All samples have informed consent and were anonymized. Every disease sample was tested positive for the specified disease before rendering to ASU. Bordetella pertussis samples were provided by Seracare Life Sciences (Seracare). Tuberculosis from University of Texas at El Paso (UTEP). Malaria from Seracare. HIV from Creative Testing Solutions (CTS). Flu from BioreclamationlVT. Dengue from UTEP. WNV from CTS. VF from Sonora Lab. Chagas from CTS. Lyme from Seracare. Hepatitis B from CTS, Syphilis from Seracare.
- Bordetella pertussis, Lyme, Syphillis, Tuberculosis, Dengue, Flu, Hepatitis B, HIV and WNV samples are used in the bacterial versus viral experiment. Chagas, Malaria and Valley Fever were added in the bacterial versus non-bacterial experiment. All samples are randomly assigned into training, validation and test set with equal probability.
- Serum samples were diluted 1:1500 into the sample buffer (3% BSA in 1×PBST) before incubated on immunosignature microarrays at a final volume of 150 ul for 1 h at 37° C. with rotating. Primary antibodies from the serum were then washed with 1×PBST for 3 times and rinsed with ddH2O for 3 times. 4 nM Secondary anti-human IgG antibodies with Alexa-Fluor 555 conjugation from Life Technologies are added in secondary incubation buffer (0.75% Casein in 1×PBST with 0.05% Tween20) to detect primary antibody binding. Secondary antibodies were incubated on the array for 1 h at 37° C. before washed off with blocking buffer. Slides were then washed with 1×PBST and ddH2O before drying. Images were obtained from scanning arrays at 555 nm using Innoscan 910 scanner. Signal intensity for features were extracted using GenePix Pro 6.0.
- Analysis is performed using scripts written in R or the JMP software (SAS Institute Inc.). Raw intensity reads for all samples are normalized to the median per sample. Quality Control (QC) for the samples is performed by checking each sample's average correlation against all other samples. Samples with correlation <0.2 are deleted. 226 samples are run on immunosignature and 212 samples passed QC and were analyzed.
- Feature selection is done by using samples in the training and validation set. Two-tail Student's T-Test is performed for each peptides by comparing bacterial infection samples versus viral infection samples (non-bacterial infection samples). Cutoff is controlled at allowing 1 false positive for all test, which is 1/124,000 or 1000 peptides, whichever is smaller.
- PCA is performed using selected peptides with all samples, with the test set samples highlighted in right PCA plot. Hierarchical clustering is performed using the selected peptides with all samples. Ward method is used in calculating the distance between the samples. The same method is used in calculating distance for the features in two-way clustering.
- Random Forest is carried out with
maximum 100 trees in the forest. Minimum split per tree is set at 10 and maximum at 2000. Early stopping rule is applied on validation set. And performance of the classifier is evaluated and output as confusion matrix for the training, validation and test set. Neural Network is built with one hidden layer and 3 nodes, with Tan H as the activation function. - Stepwise regression for reducing number of features is used with stopping rule of p-value cutoff at 0.1 for both entering and leaving the model. The model starts empty with no feature. Features become included in the model if below cutoff p-value and will be removed from the model once p-value larger than the cutoff. This process is done recursively until the model stabilize, with no feature entering and leaving the model. Then the selected features are tuned to maximum RSquare for the validation set. Then Logistic regression is used in building model with the 2 selected peptides.
- Blast search of the 2 peptides is done using the NCBI blast server. Protein Blast (blastp) suite is used. Database is Reference proteins and organism is limited to Bacteria (taxid:2). Algorithm parameters is set to adjust for short sequences, and max target sequences at 100. Then the matched sequences are processed to contain only linear matched part. The 100 matched sequences are imported into MEME suite to identify epitopes, with configurations of 10 minimum sites per epitope and 3 maximum epitopes.
- While the preferred embodiments have been illustrated in detail, it should be apparent that modifications and adaptations to those embodiments may occur to one skilled in the art without departing from the scope of the present claims.
Claims (1)
1. An array comprising at least two peptides that are capable of differentially binding to one or more antibodies produced in response to a bacterial infection or to one or more antibodies produced in response to a viral infection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US17/806,242 US20220341930A1 (en) | 2017-01-31 | 2022-06-09 | Diagnostic to distinguish bacterial infections |
Applications Claiming Priority (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US201762452825P | 2017-01-31 | 2017-01-31 | |
PCT/US2018/016185 WO2018144571A2 (en) | 2017-01-31 | 2018-01-31 | Diagnostic to distinguish bacterial infections |
US201916480887A | 2019-07-25 | 2019-07-25 | |
US16/896,117 US11360086B2 (en) | 2017-01-31 | 2020-06-08 | Diagnostic to distinguish bacterial infections |
US17/806,242 US20220341930A1 (en) | 2017-01-31 | 2022-06-09 | Diagnostic to distinguish bacterial infections |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US16/896,117 Continuation US11360086B2 (en) | 2017-01-31 | 2020-06-08 | Diagnostic to distinguish bacterial infections |
Publications (1)
Publication Number | Publication Date |
---|---|
US20220341930A1 true US20220341930A1 (en) | 2022-10-27 |
Family
ID=63041025
Family Applications (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US16/480,887 Active US10712342B2 (en) | 2017-01-31 | 2018-01-31 | Diagnostic to distinguish bacterial infections |
US16/896,117 Active US11360086B2 (en) | 2017-01-31 | 2020-06-08 | Diagnostic to distinguish bacterial infections |
US17/806,242 Abandoned US20220341930A1 (en) | 2017-01-31 | 2022-06-09 | Diagnostic to distinguish bacterial infections |
Family Applications Before (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US16/480,887 Active US10712342B2 (en) | 2017-01-31 | 2018-01-31 | Diagnostic to distinguish bacterial infections |
US16/896,117 Active US11360086B2 (en) | 2017-01-31 | 2020-06-08 | Diagnostic to distinguish bacterial infections |
Country Status (2)
Country | Link |
---|---|
US (3) | US10712342B2 (en) |
WO (1) | WO2018144571A2 (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018144571A2 (en) * | 2017-01-31 | 2018-08-09 | Arizona Board Of Regents On Behalf Of Arizona State University | Diagnostic to distinguish bacterial infections |
US12025615B2 (en) | 2017-09-15 | 2024-07-02 | Arizona Board Of Regents On Behalf Of Arizona State University | Methods of classifying response to immunotherapy for cancer |
GB201910730D0 (en) * | 2019-07-26 | 2019-09-11 | Randox Teoranta | Bovine Pathogen Array |
EP4038222A4 (en) | 2019-10-02 | 2023-10-18 | Arizona Board of Regents on behalf of Arizona State University | Methods and compositions for identifying neoantigens for use in treating and preventing cancer |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10712342B2 (en) * | 2017-01-31 | 2020-07-14 | Arizona Board Of Regents On Behalf Of Arizona State University | Diagnostic to distinguish bacterial infections |
Family Cites Families (38)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040031072A1 (en) * | 1999-05-06 | 2004-02-12 | La Rosa Thomas J. | Soy nucleic acid molecules and other molecules associated with transcription plants and uses thereof for plant improvement |
US20080260763A1 (en) * | 2004-07-01 | 2008-10-23 | The Regents Of The University Of California | High Throughput Proteomics |
US8088976B2 (en) | 2005-02-24 | 2012-01-03 | Monsanto Technology Llc | Methods for genetic control of plant pest infestation and compositions thereof |
US9732131B2 (en) | 2006-02-27 | 2017-08-15 | Calviri, Inc. | Identification and use of novopeptides for the treatment of cancer |
EP1994181A4 (en) | 2006-02-27 | 2010-05-19 | Univ Arizona | Identification and use of novopeptides for the treatment of cancer |
CA2666507A1 (en) | 2006-10-16 | 2008-04-24 | The Arizona Board Of Regents, A Body Corporate Of The State Of Arizona A Cting For And On Behalf Of Arizona State University | Synthetic antibodies |
WO2009126718A2 (en) | 2008-04-08 | 2009-10-15 | Arizona Board Of Regents And On Behalf Of Arizona State University | Tumor-associated biomarkers from the dark proteom |
US20140128280A1 (en) | 2008-04-23 | 2014-05-08 | Arizona Board of Regents, a body corporate of the State of Arizona, acting for and on behaif of Ariz | Synthetic Antibodies |
US20120021967A1 (en) | 2008-04-23 | 2012-01-26 | Arizona Board of Regents, A Body Corporate of the State Of Arizona Acting for and Behalf of Arizona | Synthetic antibodies |
US20120094271A1 (en) | 2008-09-04 | 2012-04-19 | Arizona Board Of Regents For And On Behalf Of Arizona State University | Selecting for cooperatively interacting molecules |
WO2010028214A2 (en) | 2008-09-04 | 2010-03-11 | Arizona Board Of Regents Acting For And On Behalf Of Arizona State University | Screening method to identify molecules having the ability to modulate the activity of a catalyst |
WO2010059958A2 (en) | 2008-11-20 | 2010-05-27 | Arizona Board Of Regents For And On Behalf Of Arizona State University | Artificial lectin mimics and uses thereof |
US20120065123A1 (en) | 2009-03-24 | 2012-03-15 | Johnston Stephen A | Synthetic Antibodies |
US20120190566A1 (en) | 2009-03-25 | 2012-07-26 | Susan Lindquist | Compositions and methods for inhibiting biofilms |
EP2443459B1 (en) | 2009-06-19 | 2018-12-26 | The Arizona Board of Regents, A Body Corporate Of the State of Arizona acting for and on behalf Of Arizona State University | Compound arrays for sample profiling |
WO2011029008A2 (en) | 2009-09-04 | 2011-03-10 | Arizona Board Of Regents Acting For And On Behalf Of Arizona State University | Synbodies to akt1 |
WO2012134416A2 (en) | 2009-11-18 | 2012-10-04 | Arizona Board Of Regents Acting For And On Behalf Of Arizona State University | Peptide ligands |
US20120316078A1 (en) | 2011-06-08 | 2012-12-13 | Sykes Kathryn F | Methods and materials for producing polypeptides in vitro |
WO2013067160A2 (en) | 2011-11-02 | 2013-05-10 | Arizona Board Of Regents, A Body Corporate Of The State Of Arizona, Acting For And On Behalf Of Arizona State University | Anti-bacterial polypeptides and pathogen specific synthetic antibodies |
US9482666B2 (en) | 2011-11-22 | 2016-11-01 | Arizona Board Of Regents, A Body Corporate Of The State Of Arizona, Acting For And On Behalf Of Arizona State University | High throughput selection of specific cell binding and lytic polypeptides |
CN104781668B (en) | 2012-08-29 | 2018-07-24 | 亚利桑那州评议委员会,亚利桑那州法人团体,代理和代表亚利桑那州立大学 | Immune characteristic is analyzed:Lead to the approach of early diagnosis and health monitoring |
WO2014039718A1 (en) | 2012-09-05 | 2014-03-13 | Arizona Board of Regents, a body Corporate of the State of Arizona, Acting for and on Behalf of Ariz | Methods for discovering therapeutic targets |
US10046293B2 (en) | 2012-10-17 | 2018-08-14 | Arizona Board Of Regents, A Body Corporate Of The State Of Arizona | In situ chemical patterning |
US9766239B2 (en) | 2013-03-13 | 2017-09-19 | Arizona Board Of Regents On Behalf Of Arizona State University | Synbodies for detection of human norovirus |
WO2014144383A1 (en) | 2013-03-15 | 2014-09-18 | Arizona Board Of Regents, A Body Corporate Of The State Of Arizona, Acting For And On Behalf Of Arizona State University | Non-covalent patterned chemical features and use thereof in maldi-based quality control |
US10006919B2 (en) | 2013-03-15 | 2018-06-26 | Arizona Board Of Regents On Behalf Of Arizona State University | Peptide array quality control |
US10011649B2 (en) | 2013-08-26 | 2018-07-03 | Arizona Board Of Regents On Behalf Of Arizona State University | High affinity synbodies for influenza |
WO2015095136A1 (en) | 2013-12-17 | 2015-06-25 | Arizona Board Of Regents On Behalf Of Arizona State University | Immunosignature based diagnosis and characterization of canine lymphoma |
US9757472B2 (en) | 2014-07-28 | 2017-09-12 | Arizona Board Of Regents On Behalf Of Arizona State University | Proteas stabilized antibacterial peptides for S. aureus |
US10900975B2 (en) | 2015-05-12 | 2021-01-26 | Arizona Board Of Regents On Behalf Of Arizona State University | Systems and methods of epitope binning and antibody profiling |
WO2017059239A1 (en) | 2015-10-02 | 2017-04-06 | Arizona Board Of Regents On Behalf Of Arizona State University | Methods for processing biopolymeric arrays |
US10125167B2 (en) | 2015-12-04 | 2018-11-13 | Arizona Board Of Regents On Behalf Of Arizona State University | Compositions and methods for treating MRSA infections and for sensitizing MRSA to beta-lactam antibiotics |
US10416174B1 (en) | 2016-10-26 | 2019-09-17 | Arizona Board Of Regents On Behalf Of Arizona State University | Diagnosis and prognosis for chronic fatigue syndrome |
JP2020522477A (en) | 2017-06-02 | 2020-07-30 | アリゾナ ボード オブ リージェンツ オン ビハーフ オブ アリゾナ ステート ユニバーシティ | General purpose cancer vaccine and methods of making and using same |
US20200276285A1 (en) | 2017-06-02 | 2020-09-03 | Arizona Board Of Regents On Behalf Of Arizona State University | A method to create personalized cancer vaccines |
US20190234962A1 (en) | 2017-08-31 | 2019-08-01 | Arizona Board Of Regents On Behalf Of Arizona State University | Entropy of immune health |
US12025615B2 (en) | 2017-09-15 | 2024-07-02 | Arizona Board Of Regents On Behalf Of Arizona State University | Methods of classifying response to immunotherapy for cancer |
US20210164991A1 (en) | 2018-01-19 | 2021-06-03 | Arizona Board Of Regents On Behalf Of Arizona State University | Induced common antibody response |
-
2018
- 2018-01-31 WO PCT/US2018/016185 patent/WO2018144571A2/en active Application Filing
- 2018-01-31 US US16/480,887 patent/US10712342B2/en active Active
-
2020
- 2020-06-08 US US16/896,117 patent/US11360086B2/en active Active
-
2022
- 2022-06-09 US US17/806,242 patent/US20220341930A1/en not_active Abandoned
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10712342B2 (en) * | 2017-01-31 | 2020-07-14 | Arizona Board Of Regents On Behalf Of Arizona State University | Diagnostic to distinguish bacterial infections |
US11360086B2 (en) * | 2017-01-31 | 2022-06-14 | Arizona Board Of Regents On Behalf Of Arizona State University | Diagnostic to distinguish bacterial infections |
Also Published As
Publication number | Publication date |
---|---|
WO2018144571A3 (en) | 2018-11-08 |
US10712342B2 (en) | 2020-07-14 |
US20190361021A1 (en) | 2019-11-28 |
US11360086B2 (en) | 2022-06-14 |
WO2018144571A2 (en) | 2018-08-09 |
US20210041436A1 (en) | 2021-02-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11360086B2 (en) | Diagnostic to distinguish bacterial infections | |
Merkley et al. | Applications and challenges of forensic proteomics | |
Maertzdorf et al. | Toward a unified biosignature for tuberculosis | |
Filbin et al. | Plasma proteomics reveals tissue-specific cell death and mediators of cell-cell interactions in severe COVID-19 patients | |
Khan et al. | Plasma antibody profiles as diagnostic biomarkers for tuberculosis | |
Garay-Baquero et al. | Comprehensive plasma proteomic profiling reveals biomarkers for active tuberculosis | |
Jacobsen et al. | Novel strategies to identify biomarkers in tuberculosis | |
Kamath et al. | Antibody epitope repertoire analysis enables rapid antigen discovery and multiplex serology | |
Cohen Freue et al. | Computational biomarker pipeline from discovery to clinical implementation: plasma proteomic biomarkers for cardiac transplantation | |
JP2007534086A (en) | Pattern recognition method for diagnosis of systemic autoimmune disease | |
Zak et al. | A prospective blood RNA signature for tuberculosis disease risk | |
Koyuncu et al. | CXCL1: A new diagnostic biomarker for human tuberculosis discovered using Diversity Outbred mice | |
Alves et al. | Rapid classification and identification of multiple microorganisms with accurate statistical significance via high-resolution tandem mass spectrometry | |
EP2614367A2 (en) | A method for identifying protein patterns in mass spectrometry | |
Peri et al. | Evolving serodiagnostics by rationally designed peptide arrays: The Burkholderia paradigm in Cystic Fibrosis | |
Pitarch et al. | Seroprofiling at the Candida albicans protein species level unveils an accurate molecular discriminator for candidemia | |
WO2002008755A2 (en) | Identifying antigen clusters for monitoring a global state of an immune system | |
Egertson et al. | A theoretical framework for proteome-scale single-molecule protein identification using multi-affinity protein binding reagents | |
US20220084636A1 (en) | Machine learning analysis for metabolomics classification and biomarker discovery | |
Tsai et al. | Identification of endogenous peptides in nasal swab transport media used in MALDI-TOF-MS based COVID-19 screening | |
CN111836906A (en) | Classifier for identifying robust sepsis subtypes | |
US20190064177A1 (en) | Incorporating competitive binding information into molecular array analysis to generate function networks of interactions | |
Lalvani et al. | Recent advances in diagnostic technology: applications in autoimmune and infectious diseases | |
Selvaraj et al. | BacHbpred: Support Vector Machine Methods for the Prediction of Bacterial Hemoglobin‐Like Proteins | |
Yang et al. | Unravelling potential biomarkers for acute and chronic brucellosis through proteomic and bioinformatic approaches |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: ARIZONA BOARD OF REGENTS ON BEHALF OF ARIZONA STATE UNIVERSITY, ARIZONA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:JOHNSTON, STEPHEN;WANG, LU;SIGNING DATES FROM 20180203 TO 20180411;REEL/FRAME:060192/0966 |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: NON FINAL ACTION MAILED |
|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |