Cyclome930: Large-scale replica-exchange dynamics of 930 cyclic peptide reveal thermal stability and critical metal-binding likelihood

Cyclic peptides provide low entropy, thermostable scaffolds resistant to proteolytic degradation. Till date, systematic analysis and prediction of residue-level reorganization to distribute thermal stress is limited by fragmented data resources, and the lack of cyclicity-aware predictive frameworks. Here, we present Cyclome930, the largest curated and consistently featurized resource of cyclic peptides to date, integrating 930 experimentally annotated cyclic peptides from four independent repositories with associated topology, sequence, structural coordinates, and organismal metadata, thereby expanding annotated cyclic peptide coverage by approximately 3.4-fold. We provide a novel cyclic sequence alignment algorithm that explicitly accounts for rotational symmetry and knot topology, enabling more accurate scoring of sequence similarity than conventional linear alignments. To characterize peptide thermo-stability, we performed an exhaustive all-atom replica-exchange molecular dynamics simulations (100 ns REMD; 298–400 K) for all 930 cyclic peptides. REMD trajectories reveal temperature-dependent conformational transitions and thermal-stress tensor evolution. These physics-derived descriptors inform a machine learning model to predict cyclic peptide melting points from sequence and topology. We leverage this knowledgebase to build CritiCL, a multi-classifier model to predict critical mineral metal-binding likelihood across Cyclome930. All data/ tools are available open-source (cyclome930.structf.studio) serving as a foundational resource and cloud-workbench for computational design of use-inspired cyclic peptide libraries.