Authorea

Ethan Hain edited Abstract.tex over 8 years ago

Commit id: a728b22f3a09272bb389e10050e5c63f6bd84c9a

deletions | additions

Abstract Molecular shape is a useful tool for identifying small-molecules for therapeutics. Shape similarity can be found through both alignment methods, a computationally demanding approach, or through feature vector methods, computationally less demanding. Maximization of both volume overlap and various feature points makes alignment methods computationally intensive. The computational burden of alignment can be reduced by pre-aligning shapes. The recently published Volumetric-Aligned Molecular Shapes (VAMS) search pre-aligns the entire molecule, and was shown to be effective as a pre-screen. The introduced method, fragment oriented molecular shape (FOMS) search, pre-aligns shapes to only a part of the shape in detail (i. e. the binding site). This eliminates the burden of aligning shapes, so large libraries consisting of millions of shapes are searched on the scale of a few seconds, at the loss of accuracy. FOMS has outperformed VAMS for simple protein-ligand interactions. FOMS’ performance for complex protein-ligand interactions is also explored.