Adaptive Substructure Sampling for Molecules with Flexible Alkyl Substituents

Abstract The exponential growth of conformational space in molecules with flexible side chains presents a fundamental bottleneck for atomistic sampling methods. A large fraction of this complexity arises from weakly coupled, quasi-degenerate torsional degrees of freedom that contribute little to chemically meaningful structural diversity while dramatically increasing computational cost. We present a substructure-based conformational sampling framework that separates flexible side chains from the molecular core at the graph level. Detached substituents are replaced by capped surrogate groups, enabling independent sampling of subsystems and systematic reconstruction of approximate full-molecule ensembles. A limiting variant neglects explicit sampling of the side chains and assumes extended reference conformations. This approximation is rationalized using high-level electronic-structure data for linear alkyl chains, which show that folded conformations become thermodynamically competitive only beyond well-defined lengths, particularly in solution. The approach is implemented within CREST and reduces the effective dimensionality of conformational sampling for molecules with extended flexible substituents.