How to Assess Drug Target Druggability: Evidence, Methods, and Limitations

Identifying a gene or protein associated with disease is only the first step. Before a drug discovery program commits significant resources, a harder question has to be answered: can this target actually be modulated by a small molecule, biologic, or other therapeutic modality in a way that produces a useful effect? That question is what druggability assessment addresses.

The term "druggability" is used loosely in the literature, so it is worth being precise. In the narrowest sense, it refers to whether a protein can be bound by a drug-like small molecule with sufficient affinity and selectivity. In practice, the assessment has expanded to cover multiple dimensions: structural accessibility, chemical tractability, biological validation, clinical precedent, and safety liabilities. Each layer adds evidence and complexity.

---

Starting Point: Structural and Chemical Tractability

The most established framework for small-molecule druggability comes from structural biology. A protein is considered tractable if it contains a binding pocket of suitable size, shape, and chemical character to accommodate a small molecule. The original computational work by Hopkins and Groom (2002) estimated that roughly 10% of the human proteome might be druggable in this narrow structural sense, though later analyses with improved methods have revised that estimate upward [1].

Pocket detection tools such as DoGSiteScorer, fpocket, and SiteMap identify and score candidate binding sites from crystal structures or homology models. They output metrics including pocket volume, depth, hydrophobicity, and druggability scores. These are useful filters but imperfect predictors: a pocket that scores well computationally may be too flexible in solution to bind a drug reliably, or may be occupied by a large protein-protein interface that is hard to disrupt.

Protein-protein interactions (PPIs) are a category where structural druggability has historically been underestimated. PPIs were long considered undruggable because their interfaces are large and flat. The development of venetoclax (targeting BCL-2) and MDM2 inhibitors demonstrated that "hot spot" residues concentrated within larger interfaces can be addressable, but PPI targets still require careful structural characterization before tractability can be assumed [2].

For targets without an experimental structure, AlphaFold2 and ESMFold have dramatically expanded structural coverage of the proteome. These predicted structures can be used as inputs for pocket detection, with the caveat that predicted structures may not capture conformational dynamics or ligand-induced binding sites [3].

---

Chemical Space Evidence: What Has Already Been Found

A target that has yielded active chemical matter in screening, even without a drug candidate, provides meaningful tractability evidence.

ChEMBL and PubChem aggregate bioactivity data across millions of compound-target pairs from the published literature and patents. Querying these databases for a target of interest reveals whether chemical probes, tool compounds, or early-stage leads have been reported, and at what potency. A target with multiple chemotypes showing activity below 1 µM is substantially de-risked compared to one with no reported binders [4].

The Structural Genomics Consortium (SGC) and similar initiatives have made chemical probe development a public resource. Where a high-quality probe exists for a target, it provides direct evidence of chemical tractability and can be used to validate the target biologically before internal medicinal chemistry begins [5].

One important distinction: binding is not the same as functional modulation. A compound that occupies a pocket without affecting the protein's activity (a non-functional binder) provides tractability evidence but limited biological validation. The assessment needs to track both.

---

Biological and Genetic Validation

Druggability is not only a structural question. A target needs to be causally connected to disease, not merely associated with it. This is where genetic and functional evidence becomes central.

Human genetic data is among the strongest validation available. Targets with loss-of-function or gain-of-function variants in human populations that produce disease-relevant phenotypes provide direct evidence that the target is biologically relevant in humans. The Open Targets platform integrates GWAS, rare variant, and functional genomics data to score target-disease associations across multiple evidence types [6].

CRISPR knockout and knockdown screens have scaled genetic validation considerably. Genome-wide CRISPR loss-of-function screens in cell lines and organoid models can identify genes essential for disease-relevant phenotypes. Perturb-seq, which combines CRISPR perturbation with single-cell transcriptomics, goes further by characterizing the downstream transcriptional consequences of target modulation [7].

Mouse genetic models remain standard for in vivo validation, though translational failure remains common when mouse phenotypes do not reflect human disease biology. Human genetic evidence is generally considered more predictive of clinical success than mouse data alone [8].

---

Clinical Precedent and Target Class

The clearest signal that a target is druggable is an approved drug acting on it. The human proteome can be roughly stratified by how well-studied each protein class is:

• Well-studied target classes with many approved drugs: GPCRs, kinases, nuclear hormone receptors, ion channels, proteases
• Emerging classes with growing but still limited clinical precedent: epigenetic regulators (BET bromodomains, HDACs), E3 ligases, RNA-binding proteins
• Historically challenging classes: transcription factors, intrinsically disordered proteins, scaffold proteins

This stratification matters for resource allocation. A novel kinase in a validated pathway carries less tractability risk than a transcription factor with no structural data and no reported binders, even if the transcription factor has stronger disease genetics.

Targeted protein degradation (PROTACs, molecular glues) has meaningfully expanded the druggable space by removing the requirement for a functional binding pocket. A protein that cannot be inhibited by occupying an active site may still be degradable if a surface patch can recruit an E3 ligase. PROTAC programs against targets like BRD4, AR, and STAT3 have demonstrated this, though the modality introduces its own selectivity and pharmacokinetic challenges [9].

---

Safety Liabilities and Anti-Targets

Druggability assessment is incomplete without a safety filter. A target that can be bound and that modulates disease biology may still fail if its inhibition produces unacceptable on-target toxicity.

Key checks at this stage include:

• Tissue expression profiling using GTEx and the Human Protein Atlas to identify where the target is expressed. High expression in critical tissues (heart, CNS, liver) flags potential liability.
• Paralog analysis to assess whether closely related proteins are likely to be co-inhibited by a small molecule, and whether those paralogs have known essential functions.
• Mouse knockout phenotypes from the International Mouse Phenotyping Consortium (IMPC) provide in vivo safety signals, though again with the caveat of cross-species translation.
• Human genetic evidence of target modulation in the other direction: are there people with loss-of-function variants at the target locus, and are they healthy? Naturally occurring human knockouts with no major adverse phenotype substantially de-risk the target [10].

---

Limitations of Current Approaches

A candid druggability assessment acknowledges where the methods fall short.

Structural methods miss dynamics. Crystal structures and even predicted models are static. Cryptic binding sites that open only transiently in solution, which have proven important for targets like KRAS G12C, are not captured by standard pocket detection [11].

Bioactivity databases are biased. ChEMBL and PubChem reflect what has been screened and published, which skews toward well-funded target classes and away from genuinely novel biology. Absence of reported binders is not strong evidence of intractability.

Genetic validation does not guarantee pharmacological success. A target implicated by GWAS may be relevant to disease susceptibility over a lifetime without being an effective intervention point once disease is established. The timescale and context of genetic perturbation often differ from what a drug produces.

Druggability is modality-dependent. A target that is undruggable by small molecules may be tractable by a biologic, a PROTAC, an antisense oligonucleotide, or a cell therapy. Framing druggability as a single binary question is increasingly a category error.

---

Putting It Together

A rigorous druggability assessment layers multiple evidence types: structural pocket data, chemical bioactivity precedent, genetic and functional disease validation, target class context, and safety flags. No single data source is sufficient, and the weight placed on each depends on the therapeutic area and the modality under consideration.

The practical output of this exercise is not a yes/no answer but a risk-stratified confidence level, informing whether a target warrants exploratory chemistry, a full hit-finding campaign, or a decision to deprioritize. Getting that stratification right at the outset is one of the highest-leverage decisions in drug discovery.

---

References

1. Hopkins AL, Groom CR. The druggable genome. Nature Reviews Drug Discovery. 2002;1:727–730. https://doi.org/10.1038/nrd892
2. Arkin MR, Tang Y, Wells JA. Small-molecule inhibitors of protein-protein interactions: progressing toward the reality. Chemistry & Biology. 2014;21(9):1102–1114. https://doi.org/10.1016/j.chembiol.2014.09.001
3. Jumper J et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596:583–589. https://doi.org/10.1038/s41586-021-03819-2
4. Gaulton A et al. ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic Acids Research. 2012;40(D1):D1100–D1107. https://doi.org/10.1093/nar/gkr777
5. Arrowsmith CH et al. The promise and peril of chemical probes. Nature Chemical Biology. 2015;11:536–541. https://doi.org/10.1038/nchembio.1867
6. Ochoa D et al. Open Targets Platform: supporting systematic drug-target identification and prioritisation. Nucleic Acids Research. 2021;49(D1):D1302–D1310. https://doi.org/10.1093/nar/gkaa1027
7. Dixit A et al. Perturb-Seq: dissecting molecular circuits with scalable single-cell RNA profiling of pooled genetic screens. Cell. 2016;167(7):1853–1866. https://doi.org/10.1016/j.cell.2016.11.038
8. Nelson MR et al. The support of human genetic evidence for approved drug indications. Nature Genetics. 2015;47:856–860. https://doi.org/10.1038/ng.3314
9. Bekes M, Langley DR, Crews CM. PROTAC targeted protein degraders: the past is prologue. Nature Reviews Drug Discovery. 2022;21:181–200. https://doi.org/10.1038/s41573-021-00371-6
10. King EA, Davis JW, Degner JF. Are drug targets with genetic support twice as likely to be approved? Revised estimates of the impact of genetic support for drug mechanisms on the probability of drug approval. PLOS Genetics. 2019;15(12):e1008489. https://doi.org/10.1371/journal.pgen.1008489
11. Ostrem JM et al. K-Ras(G12C) inhibitors allosterically control GTP affinity and effector interactions. Nature. 2013;503:548–551. https://doi.org/10.1038/nature12796