GPU-based acceleration of an RNA tertiary structure prediction algorithm.

Area of Science:

• Computational biology
• Structural bioinformatics
• RNA structure prediction

Background:

• Experimental RNA structure determination methods (X-ray crystallography, NMR) require substantial pure samples, limiting high-throughput analysis.
• Computational RNA tertiary structure prediction is gaining importance but existing algorithms are computationally intensive.
• There is a need for efficient and accelerated computational methods for RNA structure prediction.

Purpose of the Study:

• To propose and evaluate a parallelization methodology for the Fragment Assembly of RNA (FARNA) algorithm.
• To enhance the speed of computational RNA tertiary structure prediction.
• To leverage multi-core CPUs and Graphics Processing Units (GPUs) for accelerated RNA modeling.

Main Methods:

• Developed a parallelization scheme for the FARNA algorithm integrating multi-core CPUs and GPUs.
• Tested the parallelized FARNA approach on various RNA sequences.
• Measured computational speedup compared to baseline single-core CPU and multi-core CPU architectures.

Main Results:

• Achieved significant reductions in RNA structure prediction time.
• Demonstrated a speedup of up to approximately 24x compared to a single-core CPU baseline.
• Obtained an additional 12x speedup over a quad-core CPU setup by utilizing GPUs.

Conclusions:

• The proposed parallelization methodology effectively accelerates RNA tertiary structure prediction.
• The harmonious utilization of CPUs and GPUs offers a cost-effective solution for computational acceleration.
• This approach is highly beneficial for researchers needing faster RNA structure modeling on standard computing hardware.