Quartz 4

Confidence metrics are not general binding-affinity predictors

Created Modified

Summary

Confidence metrics from structural modeling and design neural networks are not calibrated, target-independent predictors of binding affinity for de novo binders (1,2). This includes pLDDT, ipTM, Rosetta scores, ProteinMPNN likelihoods, and ESM-2 log-likelihoods. They can nonetheless contain target-specific classification or ranking signal: PAE weakly correlates with antibody-antigen affinity and better separates binders from nonbinders (3), while AlphaFold3 ipTM distinguishes antibody binders on some targets (4,5). Binder classification or target-specific enrichment should therefore not be conflated with general affinity regression. The limitation of pLDDT is also broadly true of monomeric protein stability. Metrics like ipTM can quickly respond to even small changes in the conditioning representations used to drive diffusion-based structure prediction, showing how brittle such metrics are with respect to input sequences and MSAs (6).

Figures

Ref (1)

See also

1.
Li Q, Vlachos EN, Bryant P. Design of linear and cyclic peptide binders from protein sequence information. Communications Chemistry. 2025;8(1):211. Available from: https://doi.org/10.1038/s42004-025-01601-3
2.
Kosonocky CW, Abel AM, Feller AL, Cifuentes Rieffer AE, Woolley PR, Lála J, et al. Validation and analysis of 12,000 AI-driven CAR-T designs in the Bits to Binders competition. openRxiv; 2026. Available from: https://doi.org/10.64898/2026.03.03.709355
3.
Jin W, Chen X, Vetticaden A, Sarzikova S, Raychowdhury R, Uhler C, et al. DSMBind: SE(3) denoising score matching for unsupervised binding energy prediction and nanobody design. openRxiv; 2023. Available from: https://doi.org/10.1101/2023.12.10.570461
4.
Bennett NR, Watson JL, Ragotte RJ, Borst AJ, See DL, Weidle C, et al. Atomically accurate de novo design of antibodies with RFdiffusion. Nature. 2025;649(8095):183–93. Available from: https://doi.org/10.1038/s41586-025-09721-5
5.
Harvey EP, Smith JS, Hurley JD, Granados AJ, Schmid EW, Liang-Lin JG, et al. In silico discovery of nanobody binders to a G-protein coupled receptor using AlphaFold-Multimer. Nature Communications. 2026 Apr; Available from: http://dx.doi.org/10.1038/s41467-026-72093-5
6.
Maddipatla A, Rzayev A, Pegoraro M, Pacesa M, Schanda P, Marx A, et al. Inference-time optimization for experiment-grounded protein ensemble generation. 2026; Available from: https://arxiv.org/abs/2602.24007

Graph View

Backlinks