How big must a follow-up DEL be to train a high-resolution binding model?
- A DEL screen can find binders, but guiding medicinal chemistry needs a model that resolves affinity shifts from single-atom changes. That requires dense sampling around the hit and enough tight binders, which a naive diverse library does not supply.
- The authors model binding as a sum of small independent contributions, pixels on the target and antipixels on the ligand, and argue a second-generation library built around the hit can supply the missing data.
- From the affinity spread already seen in the naive library, with no structural information and no new synthesis, they estimate the complexity the follow-up library needs. They translate the number of tight binders a model needs into library size and expected yield.
💡 Points Based on the abstract
Independent, additive pixels are a convenient fiction: real SAR is full of non-additivity, conformational coupling and activity cliffs, so the size estimate is only as good as that assumption. And DEL read counts are noisy proxies for affinity. Before trusting any library-size formula, I'd compare it against simply synthesizing a focused hit-centered set off-DNA and measuring it.
📄 Paper · arXiv ↗ Using an energy pixels approach to estimate the size a 2nd Generation DEL must be to train a high-resolution binding model