Benchmarking interface confidence as a selection criterion for computational nanobody design against survivin's BIR-domain groove
- 1Department of Anesthesiology and Surgical Intensive Care Unit, Xinhua Hospital, School of Medicine and School of Biomedical Engineering, Shanghai Jiao Tong Univ
Abstract
Designed binders are routinely filtered on predicted interface confidence, which reports how self-consistently a model places an interface, not where it lies. We designed 200 nanobody loops against the survivin BIR-domain groove, co-folded 196, and checked epitope contact in every predicted complex. Confidence showed no detectable association with epitope engagement ({rho} = +0.100, 95% CI -0.04 to +0.24, n = 196); its highest and lowest quartiles failed alike (odds ratio 0.97, 95% CI 0.41 to 2.26). It tracked instead the run's other confidence estimates ({rho} = +0.61). Twelve-seed replication separated two candidates the criteria had not: one reached the complete epitope in 10 of 12 runs, the other never. Across 155 predictions the top-ranked model was no better than one drawn at random from its five (P = 0.610), though only 3 of 31 failed on every model. Nothing was assayed; nb_cand_057 is an in silico candidate only.
Related
- Benchmarking confidence estimation and rescoring for cyclic peptide-protein complex predictions
- Auditing Confidence-Calibrated Benchmarking of Single-Cell Perturbation Models
- Confidence Filtering in Video-Based Gait Assessment: Measurement Error versus Selection Shift
- RegDesign-Bench: Benchmarking computational methods for cell-type-specific regulatory DNA design
- Uncertainty-Aware Model Selection with a Calibrated Probability-Generating-Function-Based Bayesian Information Criterion
BibTeX
@misc{gately2026benchmarking,
title = {{Benchmarking interface confidence as a selection criterion for computational nanobody design against survivin's BIR-domain groove}},
author = {C. Gately and X. Guan and X. Ding},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.04.753929},
url = {https://www.biorxiv.org/content/10.64898/2026.10.04.753929v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Benchmarking interface confidence as a selection criterion for computational nanobody design against survivin's BIR-domain groove
AU - C. Gately
AU - X. Guan
AU - X. Ding
PY - 2026
DA - 2026/10/04
PB - bioRxiv
DO - 10.64898/2026.10.04.753929
UR - https://www.biorxiv.org/content/10.64898/2026.10.04.753929v1
AB - Designed binders are routinely filtered on predicted interface confidence, which reports how self-consistently a model places an interface, not where it lies. We designed 200 nanobody loops against the survivin BIR-domain groove, co-folded 196, and checked epitope contact in every predicted complex. Confidence showed no detectable association with epitope engagement ({rho} = +0.100, 95% CI -0.04 to +0.24, n = 196); its highest and lowest quartiles failed alike (odds ratio 0.97, 95% CI 0.41 to 2.26). It tracked instead the run's other confidence estimates ({rho} = +0.61). Twelve-seed replication separated two candidates the criteria had not: one reached the complete epitope in 10 of 12 runs, the other never. Across 155 predictions the top-ranked model was no better than one drawn at random from its five (P = 0.610), though only 3 of 31 failed on every model. Nothing was assayed; nb_cand_057 is an in silico candidate only.
ER -