Design of high-performance parallelized gene predictors in MATLAB.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene prediction algorithms are crucial for analyzing large DNA sequences.
• MATLAB is a convenient environment for bioinformatics but often overlooked for performance.
• Parallel computing offers a solution to accelerate computationally intensive bioinformatics tasks.

Purpose of the Study:

• To propose and implement parallel gene prediction algorithms in MATLAB.
• To evaluate the performance of these algorithms on both central processing units (CPUs) and graphics processing units (GPUs).
• To demonstrate significant speedups compared to conventional, non-parallel approaches.

Main Methods:

• Implementation of parallel gene prediction algorithms based on Goertzel's algorithm and Fast Fourier Transforms (FFTs).
• Utilizing varying degrees of parallelism on CPUs and GPUs.
• Employing MATLAB's Parallel Computing Toolbox, including MEX functions and PARFOR/GFOR constructs.

Main Results:

• A straightforward implementation could take over 4.5 hours for 15 million base pairs (bps).
• Optimized parallel implementations reduced processing time to under five minutes.
• A GPU implementation achieved processing in as little as 57 seconds.

Conclusions:

• Parallelism in MATLAB can accelerate gene prediction for very large DNA sequences by over 270 times.
• Direct MEX function access and PARFOR optimize CPU performance for Goertzel's algorithm.
• Data segmentation within GFOR is crucial for minimizing execution time on GPUs.