Integrating Colocalization-Centric Evidence into Target Prioritization

We have been comparing Open Targets with newer genetics-focused resources such as GPMap. While Open Targets excels at target prioritization by integrating genetics, biology, and therapeutic evidence, GPMap provides extensive colocalization and molecular QTL evidence that can help improve causal gene assignment at disease-associated loci.

In our experience, one of the key challenges in human genetics–driven drug discovery is moving from a GWAS signal to a high-confidence causal target. Resources that systematically integrate eQTL, pQTL, and other molecular QTL data through colocalization analyses can provide valuable mechanistic insights and help distinguish between nearby candidate genes. This is particularly important for RNA-targeted therapeutics, where selecting the correct target is often the critical decision point.

Has the Open Targets team considered incorporating large-scale colocalization-derived confidence metrics as an additional evidence layer for target–disease associations? It would be interesting to understand how such evidence might complement existing genetics-based approaches, including fine-mapping and locus-to-gene methodologies.

What do you think? I’d love to learn how others in the community are combining Open Targets with resources such as GPMap, colocalization analyses, and molecular QTL data in their target discovery workflows, and what approaches have been most effective in practice. Looking forward to hearing different perspectives and learning from the community.

Hi @Shicheng_Guo, thank you very much for your interest in the Open Targets Platform. We do integrate molQTL evidence as part of our L2G framework. Moreover, we provide the full results of molQTL colocalisation as a dataset and a widget on the credible set page. The presence of colocalisation is something that drives the L2G signal, so in many cases having a significant L2G signal means we also have a colocalisation with molQTL.

We are currently scoping the addition of a colocalisation signal as a separate evidence column, but our own results show that this evidence is too noisy, with FDR for eQTLs exceeding 60% (see our current preprint: https://www.biorxiv.org/content/10.64898/2026.04.28.721048v1). It may be more reasonable to visualise tissue-specific colocalisation, and we are currently working on this.