Descriptor-based Fitting of Structurally Diverse LPA1 Inhibitors into a Single predictive Mathematical Model

Omotuyi IO and Hiroshi Ueda

Abstract

120 structurally diverse compounds previously reported as LPA1 inhibitors have been used to derive a mathematical model based on their descriptors. The pre- and post-cross-validated correlation coefficient (R2) is 0.79168 (RMSE=0.61459) and 0.70939 (RMSE=0.72938) respectively. Principal component analysis (PCA) was also used to reduce the dimension and linearly transform the raw data. PCA results showed that nine (9) principal components sufficiently accounts for more than 98% of the variance of the dataset with a fitting mathematical equation. Our model accurately predicted ~86% of the compounds tested regardless of their structural diversities.