Collision cross section predictions using 2-dimensional molecular descriptors.

Traditional methods for deriving computationally-generated collision cross sections for comparisons with ion mobility-mass spectrometry data require 3-dimensional energy-minimized structures and are often time consuming, preventing high throughput implementation. Here, we introduce a method to predict ion mobility collision cross sections of lipids and peptide analogs important in prebiotic chemistry and other fields. Using less than 100 2-D molecular descriptors this approach resulted in prediction errors of less than 2%.

Local-feature analysis for automated coarse-graining of bulk-polymer molecular dynamics simulations.

A method for automated coarse-graining of bulk polymers is presented, using the data-mining tool of local feature analysis. Most existing methods for polymer coarse-graining define superatoms based on their covalent bonding topology along the polymer backbone, but here superatoms are defined based only on their correlated motions, as observed in molecular dynamics simulations. Correlated atomic motions are identified in the simulation data using local feature analysis, between atoms in the same or in different polymer chains. Groups of highly correlated atoms constitute the superatoms in the coarse-graining scheme, and the positions of their seed coordinates are then projected forward in time. Based on only the seed positions, local feature analysis enables the full reconstruction of all atomic positions. This reconstruction suggests an iterative scheme to reduce the computation of the simulations to initialize another short molecular dynamic simulation, identify new superatoms, and again project forward in time.