Recent years have seen an unprecedented growth of chemical databases incorporating tens of millions of available compounds (e.g., PubChem database) and up to 170 billion synthetically feasible chemicals (e.g., GDB-17 database). This �Big Data� context offers tremendous opportunities for discovering novel molecules with the desired therapeutic and safety profiles. However, current cheminformatics technologies are not capable of handling, characterizing, and virtually screening such extremely large chemical libraries. Herein, we present a proof-of-concept study of GPU-accelerated cheminformatics software capable of (i) calculating chemical descriptors (constitutional, topological and structural) for a billion molecules-large library (GDB-13 database) and (ii) identifying those compounds with the desired features (Figure 1). Algorithms were implemented using CUDA and Python. Overall, we show the ability of GPU-based virtual screening software to handle and characterize very large sets of chemicals, identify compounds with specific properties, and enable similarity searches using reference probes. We posit that GPU computing represents an effective and inexpensive architecture to develop and employ a new generation of cheminformatics methods and tools ready to process billions of compounds.

Toward virtual screening of billions of compounds using GPU-accelerated cheminformatics methods

247th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.