GPU-CASSERT

GPU-CASSERT performs high-performance similarity searches of three-dimensional (3D) protein structures to accelerate identification of common molecular substructures in structural bioinformatics.

Key Features:

• GPU-Based Implementation: Implements the CASSERT algorithm optimized for execution on GPUs and GPGPUs, demonstrated on a GeForce GTX 560Ti (384 cores, 2GB RAM).
• Two-Phase Alignment Algorithm: Uses a two-phase fragment-based alignment process that matches fragments of protein structures to detect shared molecular substructures.
• Massive Parallelization and Performance: Parallelizes both alignment phases on the GPU, achieving on average a 180-fold speed increase over the CPU-based counterpart running on a single core of an Intel Xeon E5620 (2.40GHz, 4 cores).

Scientific Applications:

• Large-Scale Database Screening: Scans databases containing tens or hundreds of thousands of protein structures to identify structural similarities.
• Structural Bioinformatics Analyses: Enables rapid comparative analysis and identification of common molecular substructures across 3D protein structures.

Methodology:

Implements the CASSERT two-phase fragment-based alignment algorithm with both phases parallelized for execution on GPUs/GPGPUs.