Hi, I have a general question that I would love to hear wide-ranging opinions on.

In transcriptomic or proteomic studies, we often produce a list of differentially expressed genes or proteins, for example, in a certain disease versus healthy controls. Various tools exist for interpreting these gene lists, e.g. GSEA, IPA, enrichR which are based on different methods that tend to be some variation of a Fisher test looking at the overlap between the test set of genes and literature gene lists from various databases. In more advanced single cell studies, we can define gene modules which co-vary across cells, and can then investigate the components of these modules using similar approaches.

What I am working on is how to connect these dysregulated pathways to the large number of existing therapeutics which could include targeted medicines (e.g. EGFR inhibitor), medicines with more broad mechanisms (e.g. glucocorticoids), and even compounds which may be over-the-counter or ancient medicines that have some biological activity.

One way we have done this so far is using the kinase perturbation gene sets in GEO (which are compiled in enrichR) which provides lists of differentially expressed genes when kinases are over-expressed or inhibited in different cell lines. This is a nice resource because many kinases are “druggable”. I know there is also the LINCS database which may be useful.

I was wondering what other ideas people have or pieces of literature people can refer me to?

I am especially interested in whether machine learning based approaches could try to connect biological pathways to drug targets in a way which relies less on literature databases and can learn relationships and patterns in the data which may not be apparent to a biologist.

To try out the RSC grant feature, I am assigning a 10 RSC grant to award to one individual who contributes their expertise to this question.

Tools to connect biological pathways to existing therapeutics?

Tools to connect biological pathways to existing therapeutics?