Generalized Correlative Index, A New Adaptable Molecular Characterization Method and Its Application in Predicting Collision Cross Sections of Singly Protonated Peptides

A novel adaptable and adjustable structure descriptor based on the distribution and interaction of 2D atom pairs is described. The new Generalized Correlative Index (GCI) was applied to a large-scale dataset of peptide collision cross sections which were previously measured using Ion Mobility Spectrometry (IMS). Based on Genetic Algorithm-Multiple Linear Regression (GA-MLR) approach, a linear model relating GCI with peptide collision cross section was constructed and tested using external validation and Monte Carlo cross-validation, the results confirmed that the GCI-based model is robust and predictive. The statistics for training and test sets are r=0.9960, q=0.9955, RMSEE=5.217, RMSCV=5.413, r_pred=0.9959, and RMSEP=5.016.