Therapeutic & Genomic Discovery

We use biological evidence to narrow the search for new therapies.

From human genetics and disease signatures to targets and repurposing.

This work connects genomic, transcriptomic, pharmacological, and clinical evidence to nominate plausible targets and prioritize existing compounds for new indications.

Human geneticsTarget identificationDrug repurposingTranscriptomics
Genomic ribbon transforming into a therapeutic target with a vertical marker
Research portfolio

Selected Works

01Nature Protocols · 2021

Billy Zeng et al.

OCTAD: an open workspace for virtually screening therapeutics targeting precise cancer patient groups using gene expression features

Read the publication↗
Precision oncologyGene expressionVirtual screening
Precision oncology

Can disease expression profiles become a practical therapeutic screen?

Figure 1OCTAD system design, therapeutic-prediction workflow, and web workspace. Figure reproduced from the publication.
Question

Can researchers screen therapies for molecularly defined cancer groups without rebuilding a complex computational stack for each study?

Approach

OCTAD integrates 19,127 patient-tissue samples across more than 50 cancer types with expression profiles for 12,442 compounds, then links reference-tissue selection, disease-signature generation, reversal scoring, and in silico validation.

Finding

The protocol turns heterogeneous public data into one reproducible workflow for ranking drugs and targets in a precisely defined patient group.

Why it matters

A shared workspace lowers the technical barrier to hypothesis generation, especially for uncommon cancers and clinically meaningful subgroups.

02BMC Medical Genomics · 2019

Benjamin S. Glicksberg et al.

Integrative analysis of loss-of-function variants in clinical and genomic data reveals novel genes associated with cardiovascular traits

Read the publication↗
Loss of functionCardiovascular traitsTarget validation
Human genetic evidence

Can naturally occurring gene disruption reveal actionable targets?

Figure 4DGAT2 loss-of-function associations, tissue-expression analyses, and in vivo effects of selective DGAT inhibitors. Figure reproduced from the publication.
Question

Can loss-of-function variants, longitudinal clinical traits, and tissue expression be combined to prioritize cardiovascular therapeutic targets?

Approach

The study linked clinical and genomic data from 10,511 BioMe participants with RNA sequencing from 600 STARNET participants across seven metabolic and vascular tissues, followed by experimental validation.

Finding

Loss-of-function variants in 433 genes were associated with major cardiovascular traits; 115 genes had concordant tissue-expression evidence, with DGAT2 providing a focused genetic and pharmacologic example.

Why it matters

Human genetic evidence can strengthen target nomination before costly therapeutic development and help connect population observations to tractable biology.

03Methods in Molecular Biology · 2019

Benjamin S. Glicksberg et al.

Leveraging Big Data to Transform Drug Discovery

Read the publication↗
Disease signaturesPerturbation dataReproducibility
Reproducible methods

What turns public molecular data into a defensible drug hypothesis?

Figure 2Step-by-step in silico drug-discovery workflow beginning with a disease of interest. Figure reproduced from the publication.
Question

Which data resources, analytic decisions, and validation steps make computational drug discovery rigorous and reproducible?

Approach

This methods chapter organizes public disease and drug resources into an explicit workflow: identify datasets, define cases and controls, generate a disease signature, query perturbation libraries, and rank compounds by reversal score.

Finding

The framework makes the full chain of evidence visible—from source-data selection through candidate ranking—while identifying the caveats that can invalidate an apparently compelling result.

Why it matters

Transparent methods make computational candidates easier to audit, reproduce, and advance toward experimental testing.

04Briefings in Bioinformatics · 2018

Khader Shameer et al.

Systematic analyses of drugs and disease indications in RepurposeDB reveal pharmacological, biological and epidemiological factors influencing drug repositioning

Read the publication↗
Drug repositioningKnowledgebaseNetwork analysis
Repurposing infrastructure

What can successful drug repositioning teach us about where to look next?

Figure 6Chemical-similarity, drug–target, drug-interaction, and functional networks constructed from RepurposeDB. Figure reproduced from the publication.
Question

Which pharmacological, biological, and epidemiological features characterize drugs that have already been successfully repositioned?

Approach

RepurposeDB standardized approved repositioning evidence for 253 drugs and 1,125 disease indications, then analyzed chemical structure, target networks, disease relationships, genetics, phenotypes, and population patterns.

Finding

Successful repositioning emerges from convergent evidence — compound similarity, target biology, disease relationships, and clinical context.

Why it matters

A structured reference set supports benchmarking and turns historical repurposing successes into testable principles for prioritizing future opportunities.

Therapeutic & Genomic Discovery
Convergent evidence narrows the search from plausible mechanism to testable intervention.