Ligand-Based Virtual Screening Using Bayesian Inference Network
Ammar Abdo, Naomie Binti Salim · ACS symposium series · 2011
The concept of molecular similarity has been widely used in rational drug design, where functionally similar molecules are sought by searching molecular databases for structurally similar molecules. In conventional 2D similarity methods, uncertainty in each stage of the similarity process is not considered and molecular features that do not relate to a particular biological activity carry the same weight as the important ones. In addition, since different methods have been found to retrieve different subsets of actives from the database, it is advisable to use several search methods where possible. A novel similarity searching approach using a Bayesian inference network (BIN) is introduced, where a database is ranked in order of decreasing probability of bioactivity. Our experiments on the MDDR database demonstrate that the BIN provided an interesting alternative to existing tools for ligand-based virtual screening, especially when the actives molecules being sought have a high degree of structural homogeneity. In such cases, the BIN substantially outperformed the conventional Tanimoto-based similarity searching system.