Ensemble Geometric Persistence: HIV-1 Env Glycan-Site Geometry and Sparse Structural-Context Uncertainty Assessment (Reproducibility Archive)

This archive provides the computational methods, structural coordinate files, source data, model outputs, and reproducibility documentation supporting an exploratory investigation of geometric persistence and prediction uncertainty at HIV-1 envelope (Env) glycan sites. The research examines whether local sequence divergence and sparse structural-context observations can improve the identification of unreliable glycan-site geometric predictions. The approach combines sequence-informed error estimation with residual information obtained from six observed glycan-site probes. Held-out evaluations included five BG505-derived Env structures and two JR-FL structures. In the original experiments, the six-probe method improved pooled dangerous-error discrimination relative to sequence-only prediction, with ROC AUC increasing from 0.704 to 0.945 in BG505 and from 0.522 to 0.761 in JR-FL. These results are exploratory. The original evaluations contained only two dangerous underestimates in BG505 and three in JR-FL. Within-structure discrimination did not consistently improve, and training-test construct overlap was identified. Additional construct-family exclusion analyses document the sensitivity of the findings to training composition and failure prevalence. The principal computational outputs were regenerated exactly from archived structural coordinates using the supplied scripts. The archive includes frozen protocols, structural data, prediction tables, provenance audits, family-exclusion analyses, and a companion manuscript draft. The study supports further investigation of sparse structural context for uncertainty assessment but does not establish reliable cross-strain site-level failure detection, glycan occupancy prediction, or protein synthesis prediction. This deposit is intended to support transparency, independent reproduction, methodological evaluation, and future extension of HIV-1 Env structural bioinformatics research. This research extends the author's previously deposited HIV-1 Env glycan-shield investigations by examining the predictive and uncertainty-assessment implications of geometric persistence, local structural constraints, and glycan-site organization.