DesertSci’c chemical similarity analysis software, BMOS, has been extended to enable fine-grained, 3D chemical similarity analysis of large scale datasets.
Our 2D graph-based chemical similarity metrics can now be augmented with 3D structure information when conformational data is provided. This is done by computing atomic RMSD comparisons for matching MOS groups, and combining these with the 2D similarity metrics using empirical penalty functions.
Validation testing showed that the new combined 2D-3D similarity methods can provide improved separation of small-molecule clusters, in even better agreement with the separation of compounds according to target class.
The latest extensions also include additional optimisation of input/output algorithms, and memory management.