Dr James Heward, VP Functional Genomics at Nucleome Therapeutics, discusses what could be the missing link in the chain of causality from patient to target to drug.  

The fundamental question that stalks every drug discovery project is “will this drug fail?” Most do. There is no single reason for more than 90% of clinical trials failing, but a recent study estimated that 77% of failures could be explained by poor biological knowledge around the target (i.e. unexpected toxicity, lack of efficacy)1. Put simply, most of the time we are unwittingly picking poor targets and lack the tools to predict this. If we are to reduce this dramatic attrition, then drug discovery programmes must be rooted in targets that we can clearly attribute as causal disease drivers. Despite numerous technical advances that can provide previously unthinkable insights into disease and drug biology, the simple concept of identifying causal genes and predicting clinical efficacy remains elusive.  

Modern technologies now routinely employed in drug discovery range from single-cell expression analyses to genome-wide CRISPR screens. Given the vast array of data that can be generated, great hope and investment have been pinned on AI/ML approaches to separate causality from correlative biology and ultimately identify new targets for efficacious drugs. Yet human genetics stands apart as one of the only approaches with the potential to predict an increased probability of drug efficacy2,3, with evidence also emerging around toxicity4,5. 

Outside rare disease and oncology, most genetic evidence comes from genome-wide association studies (GWAS), which compare the genotypes of tens of thousands of patients with healthy controls to identify variants conferring risk of, or protection from, disease. Targets encoded by genes with GWAS associations have been shown retrospectively to increase the probability of success by two- to three-fold2,3, and in areas such as autoimmunity almost every efficacious drug has turned out to have a GWAS association. However, most of those associations were reported long after the decision to prosecute the target had been taken. While this provides a powerful validation of the potential of human genetics, it does not provide a prospective blueprint to identify the next generation of targets.  

Most GWAS variants are non-coding 

Prospective identification of drug targets from GWAS is not straightforward. Only 5-10% of genetic associations directly affect coding gene function, with the majority instead acting through non-coding elements that regulate gene expression. To exploit the 90%+ of variants sitting in the non-coding genome, we must first understand the fundamental rules governing the regulatory wiring circuits that determine genome function in health and disease. All genes contain a promoter to initiate transcription but there is massive diversity in how genes use distal regulatory regions known as enhancers. Enhancers bind transcription factors in physical conjunction with promoters, but when viewed in a traditional linear sense they can be located a million base pairs or more away from the gene they regulate. When viewed in three dimensions, we see elegant structures that bring enhancers into close physical proximity to genes, boosting their expression. By using enhancers in a dynamic fashion, cells can modulate gene expression in complex patterns during differentiation or in response to external stimuli. It is by disrupting these promoter-enhancer interactions that most non-coding GWAS variants impact gene expression.  

Coding variants often have larger effects on disease risk and in some retrospective studies affected genes have shown a higher degree of support5. However, this does not necessarily make the associated genes better targets. For a typical coding variant, the effector gene is self-evident and it will exert its effect in all expressed tissues. A non-coding variant, by contrast, can subtly modulate the expression of its effector gene in a cell-type-specific fashion. Because of the linear distance in the genome sequence between enhancers and their genes, identifying the true causal gene associated with a particular variant remains difficult. False positives are commonly reported, leading to incorrect identification of the gene regulated by a disease-associated variant. This is the major challenge in unlocking the potential of non-coding genetics in discovery of drug targets with a high probability of success.   

Accurate identification of causal genes from the vast pool of non-coding associations is a huge opportunity to drive novel insights into disease and new drug targets. Subtle changes in gene expression are sufficient to confer meaningful risk or protection, placing the genes they regulate squarely as causal drivers of disease. And the cell-type specificity so often described as a challenge is also an opportunity to identify the precise cellular context through which a gene acts, providing an early insight into target biology unbiased by the existing literature. 

Improving causal gene discovery from the non-coding genome 

GWAS analyses have been dominated by statistical approaches which aim to prove that effects identified in different cohorts (disease risk, gene expression) can be connected. While this has given insights into new drug targets, many GWAS hits remain unexplained and there is a significant discordance between methodologies. For instance, a recent comparison across 882 GWAS-implicated regulatory regions found 515 gene assignments from expression-based statistical genetics and 145 from CRISPR interference screens, with only 69 pairs shared between them6. 

The critical information missing from these analyses is direct physical measurement of the interactions between enhancers and the promoters of the genes they regulate. This requires precise and scalable technology to locate 3D contacts within the intact genome as it exists in cells. Early generations of 3D genomic techniques had insufficient resolution to address this question, but an emerging new generation now makes this feasible. 

Enhancing the use of human genetics beyond target discovery 

At Nucleome we are pioneering the use of Micro Capture-C (MCC), which can resolve enhancer-promoter interactions at a single base pair level7,8 Combining this with AI/ML models and other functional genetic approaches has created a highly accurate and scalable variant-to-gene approach that can precisely identify causal genes and responsible cell types and states from non-coding genetic associations. We have used this approach to identify a new generation of inflammatory disease targets from non-coding genetics.  

Biology is rooted in genetics, and disease has many of its causal roots in genetic variation, the vast majority of which is non-coding. New tools and data to interpret this variation, combining lab-based and computational technologies, unlocks a massive pool of knowledge of disease which has been hidden in plain sight in the spaces between genes. These approaches are key to revealing the true potential of human genetics to find new targets for drugs with a much higher probability of success. This has the potential to transform both the impact and economics of our industry, turning the general expectation of failure into the possibility of success.  

About the author

James Heward is VP of Functional Genomics at Nucleome Therapeutics, where he leads the translation of non-coding genome insights into validated drug targets using base-pair-resolution 3D-genome technology. His career spans biotech and academia, with a longstanding focus on using novel omics technologies to understand the genome and drive drug discovery.

References 

  1. Filipova D, Boomgaarden W, Fox L, et al. Need for NAMs: a systematic evidence synthesis revealing over half a century of drug development failure. NAM Journal 2026;2:100119.  
  2. Nelson MR, Tipney H, Painter JL, et al. The support of human genetic evidence for approved drug indications. Nature Genetics 2015;47:856–860.  
  3. Minikel EV, Painter JL, Dong CC, Nelson MR. Refining the impact of genetic evidence on clinical success. Nature 2024;629:624–629.  
  4. Razuvayevskaya O, Lopez I, Dunham I, Ochoa D. Genetic factors associated with reasons for clinical trial stoppage. Nature Genetics 2024;56:1862–1867.  
  5. Tsepilov YA, Suveges D, Considine D, et al. The human pleiotropic map of GWAS associations and therapeutic implications. bioRxiv 2026 (preprint).  
  6. Ghatan S, Panten J, Oliveros W, et al. CRISPRi perturbation screens and eQTLs provide complementary and distinct insights into GWAS target genes. bioRxiv 2025 (preprint).  
  7. Hua P, Badat M, Hanssen LLP, et al. Defining genome architecture at base-pair resolution. Nature 2021;595:125–129.  
  8. Li H, Dalgleish JLT, Lister G, et al. Mapping chromatin structure at base-pair resolution unveils a unified model of cis-regulatory element interactions. Cell 2025;188:7175–7193.