P-Hint-Hunt: a deep parallelized whole genome DNA methylation detection tool

Area of Science:

• Epigenetics
• Genomics
• Computational Biology

Background:

• Whole genome DNA methylation detection is crucial for understanding diseases like cancer and diabetes.
• Current methods, such as mapping bisulfite-treated sequences, are often inaccurate and time-consuming.
• There is a need for efficient and accurate tools for DNA methylation analysis.

Purpose of the Study:

• To develop an accurate and efficient tool for whole genome DNA methylation detection.
• To address the limitations of existing DNA methylation analysis tools.
• To improve the speed and efficiency of analyzing large-scale epigenetics datasets.

Main Methods:

• Designed "Hint-Hunt" using complex alignment computation and Smith-Waterman dynamic programming for DNA methylation prediction.
• Developed "P-Hint-Hunt," a deep parallelized tool for whole genome DNA methylation detection on the Tianhe-2 supercomputer.
• Utilized CPU and Intel Xeon Phi coprocessors for parallel processing.

Main Results:

• P-Hint-Hunt is the first parallel tool for high-speed, large-scale DNA methylation detection.
• The tool eliminates mapping deviations caused by bisulfite treatment.
• A 48-fold speed-up was achieved with 64 threads on the Tianhe-2 supercomputer.
• Deep acceleration was observed on CPU and Intel Xeon Phi heterogeneous platforms.

Conclusions:

• P-Hint-Hunt offers significant speed-up and efficiency for whole genome DNA methylation detection.
• The parallelized tool effectively leverages heterogeneous computing platforms.
• This advancement provides a powerful solution for large-scale epigenetics research and disease association studies.