About PathoRx
PathoRx is a free educational tool for exploring how genes, biological pathways, diseases and drugs connect to one another. It is built for medical students, biology and pharmacy students, and researchers who want to see the shape of a molecular neighbourhood without writing a database query to get there.
What it does
Search a gene and PathoRx draws its interaction neighbourhood — the other proteins it is reported to bind, regulate or act alongside — and lets you expand outwards one step at a time. From any gene you can see the pathways it belongs to, the diseases it has been associated with, the drugs that target it, and the side effects recorded for those drugs.
The pathway catalogues from KEGG and Reactome are browsable as their own drawn diagrams, with drug targets marked on the map, so a mechanism can be read where it happens rather than inferred from a list of names. Disease pages rank the genes implicated in a condition by the strength and kind of evidence behind them, and show which pathways those genes cluster into.
Where the data comes from
Everything here is assembled from public research databases. Nothing is authored by hand and nothing is proprietary:
- Protein interactions — STRING (907,938 associations above a 0.4 confidence floor), BioGRID (109,379 experimentally reported interactions) and Pathway Commons (1,127,241 curated statements).
- Pathways — KEGG (78,554 relations across 372 human maps) and Reactome (139,614 participant links, with the published event hierarchy and diagram layouts).
- Disease associations — ClinVar, for reviewed clinical assertions, and the Jensen lab DISEASES resource, whose scores merge curated, experimental and text-mined evidence across 907,759 gene–disease links. The two are kept visibly distinct, because they mean different things.
- Drugs — ChEMBL, for 11,425 drugs with their molecular targets, development phase and recorded indications, and SIDER for 145,321 drug–side-effect links.
- Disease vocabulary — MONDO and the Disease Ontology, with OMIM phenotypes where a token is available.
The gene table covers 24,432 human genes. A coverage page in the app reports the loaded row count for every dataset, including the ones that are empty — the honest denominator rather than the flattering one.
What it is not
PathoRx is not medical advice and not a clinical decision support system. It is a study and research aid. Do not use it to diagnose, to choose or dose a treatment, or to make any decision about a real patient, including yourself.
The specific limits are worth knowing, because they are properties of the sources rather than bugs to be fixed:
- A link is not a mechanism. An interaction edge can mean a measured co-immunoprecipitation or a predicted functional association inferred partly from text mining. The source of every edge is shown for that reason.
- A drug linked to a disease is not necessarily a treatment for it. The underlying indication table records that a drug and a condition appeared together on a trial registration or a label, not which way the relationship runs. Primaquine appears under G6PD deficiency because it precipitates haemolysis in it.
- An association is not causation. A high text-mining score can mean two terms are frequently discussed together and nothing more.
- The processing is automated and can be wrong. This has been true in practice, not just in principle: an internal audit found drugs mis-assigned to a disease through a name-matching error. That class of mistake is found by looking, so if you see something that looks wrong, the feedback button on every page goes straight to the people who can fix it.
Who made it
PathoRx is an independent, non-commercial project. There are no accounts, no advertising and no third-party analytics or tracking. The source databases retain their own licences and terms; this site is a way of reading them together, not a republication of them.