A new simplex machine learning approach for analysis of structural chemical diversification processes. Comparison with other molecular modeling methods.

Abstract

Graphical Abstract Abstract. Metabolism represents highly organized system characterized by strong regulations satisfying the mass conservation principle. In this work, a new simplex-based simulation approach was developed to learn scaffold information on metabolic processes controlling molecular diversity from a wide set of observed chemical structures. This approach is based on iterative in silico combinations of molecular profiles using Scheffé’s mixture design. It was illustrated by cycloartane-based saponins of Astragalus genus containing one, two or three ramification chains with variable relative glycosylation levels. Competing and sequential glycosylation processes of different carbons were highlighted by the machine-learning simplex method. Comparisons between this simplex approach and other molecular modeling approaches were made to highlight advantages and limits of the new one.