Many of the proteins worth targeting with new drugs have never had their structures solved experimentally, and without a structure there is little to design against.

I build workflows that predict a protein’s 3D structure directly from its sequence, both on its own and bound to the molecules it naturally interacts with. Each prediction comes with confidence estimates and predicted binding strengths, so much of the work is deciding which models are reliable enough to build on.

The workflows run in batches on HPC clusters and are packaged so predictions can be reproduced and extended to new targets. They supply the starting structures that screening pipelines depend on.