A maximum common substructure-based algorithm for searching and predicting drug-like compounds

Motivation: The prediction of biologically active compounds is of great importance for high-throughput screening (HTS) approaches in drug discovery and chemical genomics. Many computational methods in this area focus on measuring the structural similarities between chemical structures. However, traditional similarity measures are often too rigid or consider only global similarities between structures. The maximum common substructure (MCS) approach provides a more promising and flexible alternative for predicting bioactive compounds. Results: In this article, a new backtracking algorithm for MCS is proposed and compared to global similarity measurements. Our algorithm provides high flexibility in the matching process, and it is very efficient in identifying local structural similarities. To predict and cluster biologically active compounds more efficiently, the concept of basis compounds is proposed that enables researchers to easily combine the MCS-based and traditional similarity measures with modern machine learning techniques. Support vector machines (SVMs) are used to test how the MCS-based similarity measure and the basis compound vectorization method perform on two empirically tested datasets. The test results show that MCS complements the well-known atom pair descriptor-based similarity measure. By combining these two measures, our SVM-based model predicts the biological activities of chemical compounds with higher specificity and sensitivity.