Sampling Chemical Space: Activity Cliffs, Extended Similarity, and ML Performance

Abstract

The presence of Activity Cliffs (ACs) has been known to represent a challenge for QSAR modeling. With its data high dependency, Machine Learning QSAR models will be highly influenced by the activity landscape of the data. We propose several extended similarity and extended SALI methods to study the implications of ACs distribution on the training and test sets on the model’s errors. Non-uniform ACs and chemical space distribution tends to lead to worse models than the proposed uniform methods. ML modeling on AC-rich sets needs to be analyzed case-by-case. Proposed methods can be used as a tool to study the dataset, with random and uniform splitting being the better overall data splitting alternatives.