Reproducibility package for "Dissecting protein identity, mutation effects, and disease mechanism in ESM-2 embeddings"

This reproducibility package contains the processed cohorts, statistical outputs, out-of-fold predictions, reports, figures, preregistration records, environment records, and frozen source snapshot supporting the study "Dissecting protein identity, mutation effects, and disease mechanism in ESM-2 embeddings." Large ESM-2 embedding arrays and upstream download caches are excluded; row-identity metadata, sequence inputs, model identifiers, and content fingerprints are included.