A New Ab Initio Algorithm and Theory of Structure Prediction for Flexible Biomolecules： Application to Peptides with Software and Database

Determining the structures of biomolecules is a challenging task. General computation schemes involve NP-hard problems. We present a new ab initio algorithm for peptides, a class of typical biomolecules, and establish a theoretical framework that can be generalized to other linear flexible molecules. A multilevel ensemble assembly operation, derived from statistical physics and utilizing quantum chemistry calculations, was developed. A novel scheme of ensemble acquisition with high completeness was developed for the basic units. The test results show that our algorithm is better than the scheme of structure extraction from the Protein Data Bank, which is predominantly used in biology. By implementing generalized multiscale strategies for solvent effects that integrate Rosetta and DFTB approaches, we performed calculations on an experimental peptide and provided a minimal requisite set to achieve ensemble completeness, which presents broad structural coverage under specific experimental conditions. This work will be beneficial to studies on molecular structure and function, as well as drug research and design.