WO2021014343A1 - Identification guidée par l'intelligence artificielle de médicaments repositionnés abordables pour traiter les leucémies - Google Patents

Identification guidée par l'intelligence artificielle de médicaments repositionnés abordables pour traiter les leucémies Download PDF

Info

Publication number
WO2021014343A1
WO2021014343A1 PCT/IB2020/056823 IB2020056823W WO2021014343A1 WO 2021014343 A1 WO2021014343 A1 WO 2021014343A1 IB 2020056823 W IB2020056823 W IB 2020056823W WO 2021014343 A1 WO2021014343 A1 WO 2021014343A1
Authority
WO
WIPO (PCT)
Prior art keywords
drugs
model
leukemia
treating
drug
Prior art date
Application number
PCT/IB2020/056823
Other languages
English (en)
Inventor
Uday Saxena
Venugopal KANDIMALLA
Ratnakar PALAKODETI
Sonia GAUR
Bharath Eswi NAGESH
Yadavalli ANURUPA DEVI
Badriram SISTLA
Subrahmanyam VANGALA
Original Assignee
Reagene Innovations Pvt. Ltd.
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Reagene Innovations Pvt. Ltd. filed Critical Reagene Innovations Pvt. Ltd.
Publication of WO2021014343A1 publication Critical patent/WO2021014343A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks

Definitions

  • This invention relates to novel methods of drug discovery, in particular cancer, utilizing artificial intelligence (AI).
  • the method involves a process tool to extract information from databases and publications relating to critical chemical and biological properties of drugs that are needed to identify drugs efficacious against leukemias.
  • the present invention provides a process tool to identify and develop affordable re-purposed drugs for diseases/conditions, exemplified by leukemia, and other rare diseases using Artificial Intelligence.
  • Cytotoxic drugs work by simply killing all cells, normal and cancerous cells without showing any differentiation. As a result of which they have unacceptable side effects like hair loss, body weight loss, nausea etc. The side effects are so severe that the cytotoxic drugs have to be administered in cycles, where the drugs are given to the patient and then there is a period of recuperation before another round of treatments. Often times the patients are unable to tolerate the drugs and discontinue the treatment.
  • New immunotherapies CAR-T
  • CAR-T New immunotherapies
  • these new therapies are un-affordable to most patients in India (cost is US $200,000 per treatment course). The situation is worse in children because their tolerance to anti-cancer cytotoxic drugs is even lower than adults.
  • the most frequent cancer in children is leukemia.
  • Pediatric lymphocytic leukemia (PLL) is a rare disease for which the treatments are not optimal. Thus, there is a need to identify newer safer drugs that can be safely used for treatment of leukemias.
  • the objective of the present invention is to use Artificial Intelligence (AI) to find new therapies for various diseases, especially leukemias as provided in the current application.
  • AI Artificial Intelligence
  • the present invention provides methods for identifying drug properties and drugs needed to be successfully applied to treatment leukemia and using them to identify new therapies for cancer, especially leukemia.
  • the invention relates to methods of identifying druggable properties (those properties that are critical for a drug to be successful for a particular disease) by defining both biological and chemical attributes that a drug needs to possess. Since it is not possible to manually examine all the available data from all drugs and their properties applicants have applied artificial intelligence and machine learning tools to extract the data.
  • the invention provides methods for identifying new uses for known drugs, methods that can be used to screen vast library of compounds and select those that could be suitable for leukemia, design new drugs by incorporating certain properties needed for efficacy in this disease and methods of predicting efficacy of a given drug.
  • the present invention provides a process tool to discover drugs for treating leukemia wherein the said process comprises the following steps:
  • the invention also provides a process tool:
  • Figure 1 Pictorial representation of the workflow of the prediction, identification and validation process Figure 2. Workflow of model development and compilation
  • Drug discovery is the process of finding new drug/medicines.
  • the cost of drug discovery of new drugs is a massive challenge in the pharmaceutical industry. It can cost upwards of US $2.5 billion and a decade or more to identify and test a new drug candidate. These costs have been increasing steadily over the years, and pharmaceutical scientists are constantly seeking ways to improve efficiency, save time and money, and speed up research progress.
  • AI Artificial Intelligence
  • AI utilizes the vast amount of data sets available for all marketed drugs and drug like compounds. Furthermore, the most desirable properties needed for a cancer drug using this massive data using AI can be extracted.
  • Leukemias are unique amongst other cancers in many ways. Firstly, they are not solid tumors that are localized in a particular organ. Secondly, they are present in many forms i.e. in acute and chronic malignancies. Thirdly they are a“moving” target because the excessively proliferated leukemic cells spill over into the circulation. Therefore, drug properties needed to treat this form of cancer is different. For example, the drug does not have to penetrate the solid tumor but enter single leukemic cells, therefore the physicochemical properties of the drugs have to be different. Secondly, the targets (proteins and genes) that drive uncontrolled proliferation in circulating leukemic cells are also different relative to solid tumors.
  • AI Artificial Intelligence
  • CMC Chemistry Manufacturing Control
  • AI-enabled prediction tools could improve the speed and precision of discovery and preclinical testing, opening up new research lines and enabling more competitive R&D strategies. Failure to demonstrate value of a drug compared to available therapies is a key factor undermining clinical trial progression or failure. Finding new niches of competitive advantage could reduce withdrawals and improve a drug’s success.
  • the present invention utilizes Artificial Intelligence (AI) to not only improve the speed and probability of finding new therapy for leukemia and pediatric lymphocytic leukemia but also reduce the cost.
  • AI Artificial Intelligence
  • the long term objective of this invention is to use this as AI guided platform for other diseases especially rare diseases (for example child sickle cell anemia which is prevalent in India) for which drugs are sparsely available or do not exist.
  • the present invention provides the following process tools:
  • the method collates drug properties of several drugs into a database. These include both drugs that are successful for all diseases, cancer drugs, specifically leukemia and drugs that have failed in other diseases but may be useful for cancer.
  • Drug properties as defined by the invention typically comprise structural, physicochemical, biochemical, pharmacokinetic (PK), and toxicity characteristics of a compound.
  • PK pharmacokinetic
  • compounds with favorable drug properties are optimized further to develop hits.
  • There exists numerous literature which reaffirms that control of physicochemical and other properties during compound optimization is beneficial in identifying compounds of candidate drug quality.
  • the Applicant's objective is to collate the drug properties for both the successful and failed drugs and identify properties that make a drug successful for leukemia.
  • the drug properties as defined herein further includes but not limited to
  • the drug properties as defined herein further include selected but not limited to Target affinity, Mechanism of action, Molecular Weight, Class of compound, Aqueous Solubility, Protein binding, Inhibitory concentration 50 (IC 50 ) in cell based models, Pharmacokinetic properties in two species, Dose response in efficacy models, in vivo Effective Dose (ED 50 ), Projected human dose, Route of dosing, Therapeutic Index in rodents, Maximum Tolerated Dose (MTD) in rodents, Toxicity profile in rodents and Formulation used. It is envisaged that the mechanism of action, expression profiles, toxicity and target affinity would be of significant importance in repurposing of drugs.
  • the chemical properties of the drugs that were queried comprised of aromatic bonds count, number of halogen atoms, acidic group count, basic group count, aromatic rings, aromatic bonds and the other 2-dimensional molecular descriptors like atom bond connectivity index and adjacency matrix.
  • the objective of using ML models is to push a cohort of similar data sets that contain relevant information to our problem statement and provide an output that helps conclude the hypothesis proposed by the problem statement.
  • ML models can be imagined as a mathematical function where we have dependent and independent variables. In the below example Y being dependent and x being independent.
  • ML models derive the function f(x) by looking at large number similar data sets that contain a lot of values for x & Y together. There are multiple ML algorithms that can be utilized for this purpose. In the current invention, the applicants have chosen decision trees that use a simple if- else ladder i.e. subroutine approach to represent the model / function.
  • the algorithm takes the data sets with dependent and independent variables marked upfront (supervised ML method). It attempts to create a correlation/weightage between each of the independent and the final dependent variable i.e. are the directly or indirectly proportional and how much do they contribute (as shown in the above diagram). This part is called the training process.
  • the dependent variables are numerical or categorical - depending on the hypothesis definition.
  • ML models can cater to all such data sets.
  • each individual property sets out to be a feature and their weights are calculated from within the algorithm.
  • the correlation maps to success and failure.
  • the logical consequence is finally represented through a decision tree.
  • the Decision tree serves as a mathematical function that maps complete set of drug properties (given as input) to a desired output: whether the drug is successful or not.
  • the model is compiled and fitted for optimal outcome and it will be further evaluated by validation data sets such as literature reports, in vitro validation and animal model validation that the researcher intends to use.
  • Figure 1 is a pictorial representation showing the process flow. For Example, using the above tool following is achieved:
  • Leukemia As an example, the inventors have chosen Leukemia as the target disease and AI was used to predict the success/failure of a new drug for leukemia.
  • AI was used to predict the success/failure of a new drug for leukemia.
  • such methodology can be designed and applied to any other cancer or other diseases/conditions and practiced by a person skilled in the art.
  • new drugs as mentioned here are randomly chosen drugs consisting of both known drugs for the said disease (for which drugs are predicted for) and other drugs with no known potential or record of treating the said disease;
  • Step 1 Machine learning Training input
  • Step 2 Decision tree classifier
  • the decision tree classifier creates the classification model by building a decision tree.
  • Each node in the tree specifies a test on an attribute, each branch descending from that node corresponds to one of the possible values for that attribute ( Figure 3).
  • Step 3a Featured weights/correlation
  • Step 3b Trained decision tree
  • the AI model developed in the above step was queried with below set of 31 drugs.
  • the 31 drugs chosen were not part of the training set.
  • the objective was to predict the success/failure of the queried molecules and correlate the results with the available literature if any. Further, the drugs which will be predicted as success will be validated in an in-vitro assay to confirm the results.
  • the structures of the 31 drugs belong to very diverse chemical classes. Table 10. Randomlychosen diverse set of drugs used for testing the model
  • the model is able to identify certain anti-cancer drugs that are known to be used in treatment of leukemias as well as those that are traditionally not thought to be anti-cancer but could be re purposed for leukemias. Furthermore, the utility of these non-cancer drugs was also supported by literature reports.
  • Cisplatin even though is a widely used anti-cancer drug, is not used for treatment of leukaemia’s and did not show up as a positive in our model
  • Cyclophosphamide is specifically not used as first line therapy for leukemia but only as a third linefor treatment as combination therapy for below mentioned types of leukaemia and not other types of leukaemia. Our model rated it as a failure which further validates the selectivity of our model.
  • AML Acute Myeloid Leukemia
  • CML Chronic Myelogenous Leukemia
  • ALL Acute Lymphoblastic Leukemia
  • the AI based algorithm model the applicants have developed was successful in predicting the outcomes of both known oncology and several non-oncology drugs which can be re-purposed for leukemias.
  • Several features of the model predictability are highlighted below: 1. Although some known anti-cancer drugs were flagged by the current AI model to be useful for leukemias, not all know anti-cancer were predicted by the model, which shows specificity and selectivity of the model

Abstract

Cette invention concerne de nouveaux procédés de découverte de médicaments et concerne un outil de processus permettant d'identifier des médicaments repositionnés abordables pour traiter une maladie ou un état, en particulier une leucémie. Ledit outil de processus fait appel à l'intelligence artificielle (IA) pour extraire des informations contenues dans des bases de données et des publications concernant des propriétés chimiques et biologiques critiques de médicaments qui sont efficaces contre les leucémies. Ainsi, la présente invention concerne un outil de processus permettant d'identifier et de développer des médicaments repositionnés abordables pour traiter des maladies/états, par exemple une leucémie, et d'autres maladies rares, faisant appel à l'intelligence artificielle.
PCT/IB2020/056823 2019-07-23 2020-07-21 Identification guidée par l'intelligence artificielle de médicaments repositionnés abordables pour traiter les leucémies WO2021014343A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
IN201941029655 2019-07-23
IN201941029655 2019-07-23

Publications (1)

Publication Number Publication Date
WO2021014343A1 true WO2021014343A1 (fr) 2021-01-28

Family

ID=74193869

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IB2020/056823 WO2021014343A1 (fr) 2019-07-23 2020-07-21 Identification guidée par l'intelligence artificielle de médicaments repositionnés abordables pour traiter les leucémies

Country Status (1)

Country Link
WO (1) WO2021014343A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116597939A (zh) * 2023-07-17 2023-08-15 青岛市即墨区人民医院 基于大数据的药品质量控制管理分析系统及方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080118576A1 (en) * 2006-08-28 2008-05-22 Dan Theodorescu Prediction of an agent's or agents' activity across different cells and tissue types
US20150356269A1 (en) * 2013-01-17 2015-12-10 The Regents Of The University Of California Rapid identification of optimized combinations of input parameters for a complex system
US20180082197A1 (en) * 2016-09-22 2018-03-22 nference, inc. Systems, methods, and computer readable media for visualization of semantic information and inference of temporal signals indicating salient associations between life science entities
US20190096526A1 (en) * 2017-09-26 2019-03-28 Edge2020 LLC Determination of health sciences recommendations

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080118576A1 (en) * 2006-08-28 2008-05-22 Dan Theodorescu Prediction of an agent's or agents' activity across different cells and tissue types
US20150356269A1 (en) * 2013-01-17 2015-12-10 The Regents Of The University Of California Rapid identification of optimized combinations of input parameters for a complex system
US20180082197A1 (en) * 2016-09-22 2018-03-22 nference, inc. Systems, methods, and computer readable media for visualization of semantic information and inference of temporal signals indicating salient associations between life science entities
US20190096526A1 (en) * 2017-09-26 2019-03-28 Edge2020 LLC Determination of health sciences recommendations

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116597939A (zh) * 2023-07-17 2023-08-15 青岛市即墨区人民医院 基于大数据的药品质量控制管理分析系统及方法

Similar Documents

Publication Publication Date Title
Packer Brain tumors in children
Gligorijević et al. Patient-specific data fusion for cancer stratification and personalised treatment
Chung et al. Functional Module Connectivity Map (FMCM): a framework for searching repurposed drug compounds for systems treatment of cancer and an application to colorectal adenocarcinoma
Stetson et al. Computational identification of multi-omic correlates of anticancer therapeutic response
CN109074420A (zh) 用于预测靶向药物治疗疾病的效果的系统
Nam et al. Drug repurposing with network reinforcement
Yakar et al. Prediction of radiation pneumonitis with machine learning in stage III lung cancer: a pilot study
Wang et al. A novel methodology for multicomponent drug design and its application in optimizing the combination of active components from Chinese medicinal formula Shenmai
WO2021014343A1 (fr) Identification guidée par l'intelligence artificielle de médicaments repositionnés abordables pour traiter les leucémies
Cunha et al. Adjuvant chemotherapy with mitoxantrone for cats with mammary carcinomas treated with radical mastectomy
Athar et al. Fundamental physical and chemical concepts behind “drug-likeness” and “natural product-likeness”
CN104115151B (zh) 用于鉴定具有期望生物活性的剂的方法
Gong et al. In silico prediction of potential drug‐induced nephrotoxicity with machine learning methods
Qiu et al. Artificial intelligence and cheminformatics-guided modern privileged scaffold research
Patel et al. Mutation based treatment recommendations from next generation sequencing data: A comparison of web tools
Papiez et al. Integrating Expression Data from Different Microarray Platforms in Search of Biomarkers of Radiosensitivity.
Lee et al. Development and evaluation of drugs: from laboratory through licensure to market
Li et al. Deciphering the mechanism of Indirubin and its derivatives in the inhibition of Imatinib resistance using a “drug target prediction-gene microarray analysis-protein network construction” strategy
Khayyer et al. Oral squamous cell cancer protein-protein interaction network interpretation in comparison to esophageal adenocarcinoma
MacLean et al. A Network Polypharmacological Approach to Combinatorial Drug Repurposing for Diffuse Intrinsic Pontine Glioma
MacLean et al. A network polypharmacology approach to drug repurposing for diffuse intrinsic pontine glioma
Godara et al. The application of computer-aided drug design methods for developing natural compound-based therapeutics against SARS-CoV-2
Loging The art and science of the drug discovery pipeline: history of drug discovery
CN115862890B (zh) 一种抗癌药物筛选方法及系统
Gousiadou et al. Development of QSAR ensemble models for predicting the PAMPA Effective Permeability of new, non-peptidic leads with potential antiviral activity against the coronavirus SARS-CoV-2.

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: 20844698

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20844698

Country of ref document: EP

Kind code of ref document: A1