Accelerating calculations of RNA secondary structure partition functions using GPUs

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• RNA molecules fold into complex 3D structures crucial for their diverse cellular functions.
• RNA secondary structure, determined by base-pair contacts, dictates overall conformation.
• Predicting RNA secondary structure from its sequence is computationally intensive.

Purpose of the Study:

• To accelerate the computation of RNA secondary structure base-pair probabilities.
• To leverage parallel graphics processing units (GPUs) for enhanced computational efficiency.

Main Methods:

• Implemented CUDA for calculations on GPU hardware.
• Introduced modified recursions to reduce memory usage by approximately 25%.
• Evaluated the impact of single-precision calculations on accuracy.

Main Results:

• GPU implementation reduced execution time by a factor of ~60-116 compared to serial methods.
• Single-precision calculations introduced negligible deviations in base-pair probabilities.
• Memory usage was reduced by ~25% through modified recursions.

Conclusions:

• GPUs significantly accelerate RNA secondary structure partition function computations.
• Accurate base-pair probabilities for large RNA sequences can be calculated efficiently.
• The developed GPU code is integrated into the RNAstructure software package.