Peptide structural plasticity is predictable from sequence and environment.

Biomolecular structure is commonly predicted from sequence as a single structural model, yet many molecules function through conformational ensembles that reorganize with their surroundings. Whether such structural responsiveness can be learned jointly from sequence and environmental context remains unresolved. Here, we use short peptides as an experimentally tractable system to test this principle. We assembled a large experimental multi-environment dataset for peptides' secondary structure, comprising more than 1,500 peptides and more than 5,500 peptide-environment observations across aqueous, co-solvent, and membrane-mimicking conditions. We developed ApexFold, an environment-conditioned AI framework that combines sequence representations with physicochemical descriptors of the surrounding medium to predict circular-dichroism-derived fractions of α -helical, β -like, and unstructured conformations. In two later-collected panels excluded from model development, ApexFold captured the direction and magnitude of environment-induced structural redistribution and the peptide-specific degree of plasticity, while outperforming solvent-agnostic and composition-based baselines. Static structural references, which return a single conformation, cannot represent these condition-dependent response profiles. These results establish structural responsiveness as a learnable property of sequence and environment, extending biomolecular prediction beyond static structure toward predicting-and ultimately designing-how molecules respond to the contexts in which they function.