We proposed a project that involved exploring novel chemical spaces to find inhibitors for the parasitic cysteine protease cruzipain. By using FTrees, we were able to use 11 seed compounds and filter huge chemical spaces to arrive at ~1 million compounds to perform virtual screening on. Subsequent filters for cruzipain activity are still being developed.
After 1 year, Rafael has achieved the following goals:
- We were able to filter the chemical spaces by using FTrees and find analogs by fuzzy pharmacophore similarity to 11 seed compounds with known cruzipain activity. With this, we arrived at ~1 million commercial compounds, that we needed to further filter.
- We used open access property filters (ADME, toxicity), but also developed internal filters by training machine learning models on public and in house databases. Some of these models worked well and ended up turning into different projects. We are still struggling to get a good, generalizable model for cruzipain activity, which delayed this project.
- As we are still developing the final models, we unfortunately were not able to achieve all the milestones proposed. Overall, we believe the project can still provide interesting results and we expect to purchase compounds and obtain experimental values soon. We thank BioSolveIT for offering the free licenses, which will be acknowledged when the results are published.