GPU-I-TASSER

GPU-I-TASSER accelerates I-TASSER-based protein structure prediction by using GPU-accelerated replica-exchange Monte Carlo simulations and OpenACC parallelization to enable faster tertiary-structure modeling.

Key Features:

• I-TASSER framework: Builds upon the Iterative Threading Assembly Refinement (I-TASSER) methodology used for protein tertiary-structure prediction and evaluated in CASP experiments.
• GPU acceleration: Offloads compute-intensive tasks to GPUs to substantially increase processing throughput compared to CPU-only execution.
• OpenACC parallelization: Uses OpenACC directives to parallelize core simulation routines on GPU architectures.
• Replica-exchange Monte Carlo simulations: Accelerates the replica-exchange Monte Carlo simulation component integral to I-TASSER structure assembly and refinement.
• Benchmark performance: Demonstrates an average speedup of approximately tenfold on a benchmark set of 71 protein structures versus the CPU implementation.
• Maintained accuracy: Achieves prediction accuracy comparable to the original CPU-based I-TASSER while providing accelerated performance.
• Scalability for large-scale modeling: Designed to address computational limitations of CPU-based I-TASSER for large-scale structure modeling tasks.

Scientific Applications:

• Protein tertiary-structure prediction: Produces predicted 3D structures for proteins using the I-TASSER modeling framework.
• Large-scale structure modeling: Enables high-throughput modeling campaigns by reducing compute time per target.
• Structure-based functional inference: Supports inference of protein function and interaction studies from predicted structures.
• Method benchmarking: Serves as a platform for computational performance benchmarking of structure prediction workflows.

Methodology:

Implements the I-TASSER approach with replica-exchange Monte Carlo simulations parallelized via OpenACC to exploit GPU hardware, and was empirically evaluated on a benchmark of 71 protein structures showing ~10× speedup with comparable accuracy.