Identification of and Applications in Patient Enrichment ... · Outcomes Research Epidemiology Modelling Economic Modelling Value Evidence Communication ... To prioritize cancer indications

Genes from 3 Approaches

Identification ofPredictive Biomarkersand Applications inPatient Enrichment Strategies

Case Study

The PurposeThe partner had a pipeline molecule that was under clinical development. The interest area was to

identify biomarkers indicative of drug-response in patients and further utilize the biomarkers for patient

stratification in clinical trials.

About the Client

The Excelra Approach

Client RequirementThe focus was on analysing the proprietary gene-expression data of 118 cell-lines that were treated with

the drug. Furthermore, after prediction of drug-response biomarkers, gene expression profiles of 11

patients was shared by the partner to retrospectively classify them into responders and non-responders.

Machine learning models were built using three different methods to prioritize biomarkers associated

with drug-response. Pathway enrichment analysis was performed to understand the role of the

biomarkers in disease pathophysiology. Stratification of patients based on these biomarkers resulted in

correct prediction of drug response in 8 out of 11 patients.

For 118 cell Lines: Data collection & Normalization (expression, mutation, response class)

LOCATION

USA

THERAPEUTIC AREA

Non-Hodgkin's Lymphoma

INDUSTRY

Small Pharma

COSMIC Array Express

Supervised ML ApproacheandAlgorithms (3 methods)

Supervised ML analysis

Random Forest (RF)based regression analysisto assign weighted scoreto each gene

Heat map to visualizepattern between resistantand sensitive cell lines

Partial Least Squares (PLS)method to stratify patientsinto sub-types

CCLE

Client data:IC50 values will be usedto annotate sample toDrug-XXXX sensitive/resistant class

Sensitive Set Resistant Set

Cumulative Rank

Client data:

3 Approaches of

11 patients

*Retrospective validation

Functional enrichment& assessment:

PPI, Pathways & Biological Rationale

Prediction of eachpatient’s drug

response

Match with originaldata of each patient’s

drug response

Predicted (Excelra) vs. Observed (Client)

Prioritized genes forDrug-XXXX response

8 biomarkersidentified for

drug response.

9 patients’ data werepredicted correctly

out of 11.82% prediction

accuracy.

Gene signatures used toperform sub type-levelanalysis and patient

stratification.

Transition from NHLto other tumor types.

Establish theimmune-modulatory

role and defined MOA.Opened possibilities for

combinations with IO agents.

For more information, visit https://www.excelra.com/clinical/#precision_oncology

www.excelra.com

About Excelra

Excelra's data and analytics solutions empower innovation in life sciences across the value chain from discovery to market.The Excelra Edge comes from a seamless amalgamation of proprietary curated data assets, deep domain expertise anddata science. The company's multifaceted teams harmonize and analyse large volumes of disparate unstructured data using cutting-edge technologies. We galvanize data-driven decisions to unlock operational efficiencies to accelerate drug discoveryand development. Over the past 18 years, Excelra has been the preferred data and analytics partner to over 90 global clients,including 15 of the top 20 large Pharma.

[email protected]

Excelra’s Contribution

Excelra’s Service Portfolio

Chemistry Curation Services

Biology Curation Services

GOBIOMBiomarker intelligence database

Clinical Trial Outcomes Database

RWE & Big Data Realization

SLR & Meta-analysis

GOSTARStructure Activity Relationship database

Data

Clinical

Technology

Solutions

Discovery

Translational

ValueEvidence

Target Identification

Target Dossier Services

Data Science DrivenDrug Discovery

Biomarker Discovery

Drug Repositioning

Life Cycle Management

Systems Biology Informatics

Precision Oncology Informatics

Clinical Pharmacology

Outcomes Research

Epidemiology Modelling

Economic Modelling

Value Evidence Communication

Insights

Enterprise Data Strategy

Enterprise Cloud Ops

Enterprise Digital Transformation

PortfolioAugmentation for a Potential Biologic Drug


The partner had a large molecule asset that was under development for blood cancer. They wereinterested in expanding the therapeutic potential of the molecule to solid tumours to augmentthe existing portfolio.

The focus was on leveraging public gene expression data-sets from cancer patients treated with a drugcandidate with similar mechanism of action. Machine-learning based predictive models were built usinginformation on drug responsiveness and disease gene signatures.

Predictive models were built using an iterative approach wherein patient-level and disease-level geneexpression profiles were used as input data. Clustering of cancer indications was done to prioritizeindications which were potentially sensitive to the treatment. Biological rationale was built was each ofthe prioritized indications by converging the drug mechanism of action with disease pathophysiology.

Client Requirement

About the ClientLOCATIONEurope

THERAPEUTIC AREAOncology

INDUSTRYBiotech

The Purpose

2months

3 FTE

Multiple Corroborative ML Classifiers

Public domain patientdrug response data

Disease geneexpression datasets

Disease sub-typespecific datasets

Literature genes fromother indications

Classifier-2ML

Bladder CancerCholangiocarcinoma

Stomach CancerTNBC_BL1

Sensitivepatient

Resistantpatient

Classifier-1

ML

Perturbed genes in sensitive patient

Predicted gene signature

Prioritized Indications for given drug

Disease sub type clustering

Gene centric pathway elucidation

Cancer patient data from client(known drug response)

ALL Liver cancerOvarian cancer

AML Drug Responders Potential Responders

TNBC

Non-Responders

Case Study


Created value for shareholders andthe Board to fund future programs. Potential revenue generation >$2Bil*.

Excelra facilitated portfolio optimization and expansion for the partner, by Prioritizing 10 oncology indications, a mix of solid & liquid tumors types.

Determined causal genesignatures and provided acomprehensive biological

rationale and pathway analysis.

ALL (Acute Lymphoblastic Leukemia) determined as a top-priority indication by

Excelra, was further confirmed by the client, that successfully

validated our approaches.

Portfolio enhanced fornext 2 years.

Prediction of drug-response at a

cancer subtype level.

Increased applicationof partner’s technology

platform & external validation.







SLR & Meta-analysis


Data

Clinical

Technology

Solutions

Discovery

Translational

ValueEvidence




Biomarker Discovery

Drug Repositioning





Outcomes Research


Economic Modelling


Insights





www.excelra.com

About Excelra


[email protected]

0 Resistant, 1 Sensitive

CombinationFeasibility Predictionfor Checkpoint Inhibitors for a Biologic

Case Study

The PurposeThe partner had a large molecule in the development pipeline for cancer indications. They wereinterested in combining their proprietary molecule with already approved immune check-pointinhibitors to improve therapeutic efficacy.

Client RequirementTo prioritize cancer indications based on their sensitivity towards the combination of the biologicwith a check point inhibitor (anti-PD-1/PDL-1). Publicly available data on successful and failed drug combinations was used for building predictive models.

Machine learning models were built using to assess the sensitivity of cancer indications as well aspatients to the drug combination. Based on the analysis, some cancer indications were prioritized forfurther assessment. A biological hypothesis was built to establish the synergistic role of the combinationpartners for cancer treatment.

About the ClientTHERAPEUTIC AREA

Oncology

LOCATION

Europe

INDUSTRY

Biotech

Each patient level insightOverall cancer level

ALL

Anti-PD1Sensitivity(RPART)

0 2.08 750 131 619 17.47

49.60

31.84

89.07

14.21

48.77

36.77

39.52

80.00

188

122

61

465

209

23

251

4

18557

497

77

199

13

16

164

373

179

558

542

408

415

36

20

0.55 Yes

Yes

Yes

Yes

Yes

Yes

No

No

No

No

0.31

0.29

0.93

0.72

-0.40

0.61

0.32

0.01

0.74

0.34

2.26

2.44

-0.83 -0.30

-1.05

0.41

0.88

-0.20

0

0

-1

-1

-1

-1

-1

-1

-1

Anti-PD1PLS

score

DrugX PLS score

Both (AntiPD1+

DrugX)Sensitive

TotalSample

Sensitivesample

(both DrugX+Anti PD1)

ResistantSample

ResponseRate

(Anti-PD1blocker)

LIHC

PAAD

OV

AML

BLCA

CHOL

STAD

TNBC BL1

BRCA

Col

or s

chem

e is

bas

ed o

n D

rug

X re

spon

se

RPART PLS scorePositive SensitiveNegative Resistant

Yes Must be Drug X positiveand either of Anti PD1 predictor

(RPART or PLS) as sensitive -1 Partial Sensitive


Scoring System

BioGRID

GDSC

CCLE

STITCH

TCGA

Network Based Analysis

Algorithmic-guided Screening of Drug Combinations

Drug Combinations Based on Clinical Side-effects

Based on Molecular & Pharmacological Data

Semi-supervised Learning

Mathematical Modeling of Drug-targeted Signaling Pathway

Algorithms


Feasibility/synergy predictionof the two-drug combination.

Widen the list of indicationwhere the query drug may

be developed.

Indications resistant or werepartially sensitive to the

monotherapy were predicted tobe sensitive towards combination

with the checkpoint inhibitor.

Prioritize the indication where indication therapy with PD-1 willwork the best.

Custom pathways were generatedto understand crosstalk between the

drug-induced signaling and checkpointinhibitor signaling pathways.


www.excelra.com

About Excelra


[email protected]







SLR & Meta-analysis


Data

Clinical

Technology

Solutions

Discovery

Translational

ValueEvidence




Biomarker Discovery

Drug Repositioning





Outcomes Research


Economic Modelling


Insights




Documents

Identification of and Applications in Patient Enrichment ... · Outcomes Research Epidemiology Modelling Economic Modelling Value Evidence Communication ... To prioritize cancer indications