Users appreciate researchers highlighting challenges and proposing rigorous protocols for evaluating de novo molecular binder designs because the discussion promotes better standards in the field.
No Digg Deeper questions have been answered for this story yet.
These recommendations are not meant as strict pass/fail criteria but rather as a set of best practices that, in our view, would strengthen transparency, reproducibility, and comparability across studies.
@HannesStaerk Don’t overload the chips and probes!! That is just as bad as with evaluating with dimeric/Fc fused targets and gives a lot of avidity effects which makes thing seem picomolar/low nanomolar when they are not
Two aspects in current evaluations particularly warrant scrutiny:
1. Screening signals too ambiguous to confidently call a binder are nonetheless reported as binders. 2. Affinities are computed from data that cannot support such claims.
@HannesStaerk This is great, we also went over some of proteinbase's data and some were genuinely hard to believe. People are reporting the Kd (maybe Kon and Koff) and absolutely miss mentioning that some curves looks pretty bad.
@MartinPacesa Thanks Martin, any suggestions and discussion around binder evaluations are highly appreciated 🙏
Hence, this blog post about binder evaluations and our proposal for a more rigorous standard going forward. We have adopted this practice in our updated BoltzGen and new BoltzProt-1 papers + we encourage the field to consider it! https://boltz.bio/boltzprot1-technical-report.pdf
We encountered this in our own work, where we reported binders as classified by @adaptyvbio. In follow-ups with other CROs, we found some classifications to disagree. In collaboration with @adaptyvbio, we observed that these same issues recur across published work.
@HannesStaerk Spot on! @CalebLareau
@MartinPacesa Thanks Martin!!