No, you are right. It is very hard. We can take this offline if you want to discuss. It ends up being complicated to make ML methods respect complex domains. Happy to talk though!

I think you have the right sense here. One can combine domain knowledge with ML methods!

Not happy either! It would have been better to have our offer of help accepted by Gabriele and the rest of the DiffDock authors. You, me, and @annclevesjain.bsky.social have close to 100 person-years of experience in this area. We offered our help for free!

Looks like yet another case of overhyped results due to poor #bioMLeval evaluation of deep learning models -this time deep docking methods - specifically DiffDock. Look forward to the DiffDock authors response. But dont see any major flaws in this critique. Conclusion is REALLY worth reading!

Thanks Derek! Ann (@annclevesjain.bsky.social) and I enjoyed your summary as well as your discussion of the linkage to similar observations other high-profile deep-learning reports!