Prediction\nof Synthetic Accessibility Based on Commercially Available Compound\nDatabases

Abstract

A compound’s\nsynthetic accessibility (SA) is an important aspect of drug design,\nsince in some cases computer-designed compounds cannot be synthesized.\nThere have been several reports on SA prediction, most of which have\nfocused on the difficulties of synthetic reactions based on retro-synthesis\nanalyses, reaction databases and the availability of starting materials.\nWe developed a new method of predicting SA using commercially available\ncompound databases and molecular descriptors. SA was estimated from\nthe probability of existence of substructures consisting of the compound\nin question, the number of symmetry atoms, the graph complexity, and\nthe number of chiral centers of the compound. The probabilities of\nthe existence of given substructures were estimated based on a compound\nlibrary. The predicted SA results reproduced the expert manual assessments\nwith a Pearson correlation coefficient of 0.56. Since our method required\na compound database and not a reaction database, it should be easy\nto customize the prediction for compound vendors. The correlation\nbetween the sales price of approved drugs and the SA values was also\nexamined and found to be weak. The price most likely depends on the\ntotal cost of development and other factors.