Science
PanXome utilizes ligand homology modeling to identify potential ligands for target proteins for which there are no known small molecule binders, nor known structures of the protein. Ligand homology modeling employs ligands from homologous template proteins for which structures have been determined and ligands are known. We extend this methodology to protein binding pockets for which there are no homologous holo templates. This is done by exploiting the fact that there are a limited number of distinct small molecule binding pockets across all proteins. We find a homologous binding pocket and use this as a template for virtual ligand screening.
Conventional Experimental Approach
- Conventional approach requires experimentally solving a high resolution protein crystal structure.
- Experimentally solved structures or close homology models are only available for about 25% of the human exome.
- Conventional approach is computationally inefficient and ineffective on predicted structures.
- Even if structures were available, using this approach for side effect predictions is impractical due to computational inefficiency.
PanXome Predicted Structure Approach
- PanXome utilizes homology and ligand similarity which is very effective on both experimental structures and predicted structures.
- PanXome's method is effective on 86% of the human exome.
- PanXome's methodology is extremely computationally efficient, utilizing both high resolution and low resolution predictions.
- Because of computational efficiency and span of coverage, PanXome's approach is very effective at side effect predictions.