GPU implementation of Turbo Similarity Searching using binary and fragment occurrence descriptors

Turbo Similarity Search (TSS) has been proved as one of the best simplest method for virtual screening. A lot of researches have been done to investigate the performance of TSS in different environment e.g. using MDDR database on Graphic Processing Unit GPU. However, TSS has not been tested with larger database such as ChEMBL database on similar platform. This has become the main motivation for this study to investigating the effect of different representation of chemical structures, binary and fragment occurrence descriptors towards the speed and adaptability in the architecture of GPU using TSS algorithm inside MDDR and ChEMBL database. At the end, this study gained that both parallel version of TSS using binary and fragment occurrence descriptors are scalable on the ChEMBL database. However, fragment occurrence descriptors performed better compared to binary descriptors. Therefore, this study could conclude that the complexity of the descriptors gives a bigger impact on the TSS's execution performance.