Computational target discovery with multi-omics: how to build an evidence-weighted shortlist
How biotech teams prioritize drug targets: human genetics, single-cell and spatial specificity, CRISPR dependency, tractability and safety.
Key takeaways
- Targets with human genetic support are substantially more likely to succeed clinically. Weight causal evidence highest.
- Differential expression alone mostly finds consequences of disease, not causes.
- Single-cell and spatial data tell you which cells express a target, which matters for both efficacy and safety.
- Make every evidence stream and weight visible, so the team can debate the ranking rather than trust a black box.
Choosing the target is the most consequential decision in a drug discovery program, and one of the hardest to undo. Omics data gives teams more candidate targets than they can ever test. The job of computational target discovery is to turn that abundance into a short, defensible list, with the evidence laid out so scientists, chemists and investors can challenge it.
Start with causality, not correlation
Genes that change in diseased tissue are often consequences of disease, such as inflammation, cell death or compensation, rather than causes. Human genetics offers a rare causal anchor: variants that alter a gene's function and change disease risk.
The pipeline data supports weighting this evidence highly. Nelson and colleagues found that genetically supported targets were roughly twice as likely to reach approval, and Minikel and colleagues estimated the relative success of genetically supported target-indication pairs at about 2.6 times.
Genetic evidence streams we use:
- GWAS credible sets, fine-mapping (SuSiE) and variant-to-gene assignment.
- Colocalization with eQTL and pQTL to link risk loci to a gene's expression or protein level.
- Mendelian randomization with protein QTLs to estimate the direction and size of causal effects.
- Rare-variant burden from exome cohorts, and Mendelian disease genes.
Add cell-type and tissue context
A target is only useful if it acts in the cells that drive disease.
- Disease single-cell atlases: which cell types express the target, and does expression change in disease within those cells (by pseudobulk, not cell-level tests)?
- Spatial data: are target-expressing cells in the lesion, at the invasive front, or near the immune infiltrate?
- Specificity: broad expression across many cell types and tissues is a safety flag. Highly specific expression is an opportunity.
Add functional evidence
- CRISPR screens in disease-relevant models, analyzed with MAGeCK or BAGEL2, show whether perturbing the gene produces the desired phenotype.
- Perturb-seq shows how a perturbation moves cell state toward or away from a disease signature.
- DepMap dependency data adds information on essentiality and selectivity, especially in oncology.
- Connectivity mapping (CMap, L1000) links target perturbation signatures to compounds and disease signatures.
Ask whether it is drug-able, and safe
- Tractability: known chemistry (ChEMBL), predicted pockets (fpocket on AlphaFold structures), surface accessibility for antibodies, and modality fit (small molecule, biologic, oligonucleotide, degrader).
- Safety: essential-gene status, knockout phenotypes in mouse and human, expression in critical tissues (heart, liver, brain), paralog redundancy and pharmacovigilance signals.
- Landscape: clinical trials, competitors and patents for the same target or mechanism.
Combine evidence transparently
We score each evidence stream separately, normalize, and combine them with weights the team agrees on. Every target then gets a dossier with each piece of evidence, its source and its quality.
What our experts avoid: single composite scores that nobody can interrogate. When the program team can see that a target ranks highly because of strong colocalization but weak tractability, the discussion turns to the right experiment. A black-box score only invites the question "do we trust the model?".
A typical engagement
- Define the indication and decision criteria with the program team.
- Assemble public evidence for every gene.
- Add proprietary data: your single-cell, spatial, screening or proteomics results.
- Rank and review with the team, adjust weights, and stress-test the top candidates.
- Deliver dossiers and an interactive dashboard for ongoing use.
How OmicsQuest can help
Our target discovery service runs this process for biotechs and academic drug discovery groups, hourly, as a fixed scope, or with an embedded duo team. Downstream, our drug discovery bioinformatics team supports mechanism-of-action, repurposing and biomarker work.
Frequently asked questions
Why does genetic evidence matter for drug targets?
Retrospective analyses of drug development pipelines found that targets supported by human genetic associations are more likely to progress through clinical trials to approval. Nelson et al. (2015) reported roughly a twofold enrichment, and Minikel et al. (2024) estimated a relative success of about 2.6 times for genetically supported target-indication pairs.
What data do you need to start target discovery?
You can start with public data alone (Open Targets, GWAS Catalog, GTEx, CELLxGENE, DepMap, ChEMBL) for a defined indication. Your own omics data, such as disease single-cell, spatial, CRISPR screens or proteomics, adds proprietary evidence and usually differentiates the shortlist.
Work with OmicsQuest. Related services:
References
- Nelson MR, Tipney H, Painter JL, et al. The support of human genetic evidence for approved drug indications. Nat Genet. 2015;47:856-860.
- Minikel EV, Painter JL, Dong CC, Nelson MR. Refining the impact of genetic evidence on clinical success. Nature. 2024;629:624-629.
- Li W, Xu H, Xiao T, et al. MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens. Genome Biol. 2014;15:554.
- Dixit A, Parnas O, Li B, et al. Perturb-Seq: dissecting molecular circuits with scalable single-cell RNA profiling of pooled genetic screens. Cell. 2016;167:1853-1866.
- Lamb J, Crawford ED, Peck D, et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science. 2006;313:1929-1935.
- Subramanian A, Narayan R, Corsello SM, et al. A next generation Connectivity Map: L1000 platform and the first 1,000,000 profiles. Cell. 2017;171:1437-1452.
- Szklarczyk D, Kirsch R, Koutrouli M, et al. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res. 2023;51:D638-D646.
- Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596:583-589.