This project was made by the new Scientific Interface & Tooling Lab (sitlabs.org) at Oxford. We will be releasing new interfaces and tools to improve scientific productivity with an initial focus on biology research.

Next we provided the variant mentions and paper context to gpt-4o-mini and requested structured output summarising the scientific finding. Finally we mapped all variants to the latest genome build and plotted them on our custom genome browser.

To begin we downloaded 7TB's of bioRxiv+medRxiv preprints and extracted mentions of genetic variant ID's or positions. Documented in "How to download bioRxiv on a budget": sitlabs.org/writing/bior...

This project started after I was trawling through Google Scholar to review a gene when I realised I had missed relevant upstream findings which hadn't been linked to the gene. I wondered if there could be a better interface to explore genetic literature?