An Ultrafast GPU-Enabled MGVB

Area of Science:

• Computational Biology
• Proteomics

Background:

• MGVB is a computational proteomics toolset for analyzing peptide MS/MS data.
• The original MGVB algorithm for identifying post-translational modifications was resource-intensive.
• High-performance computing clusters were required for practical application of the original algorithm.

Purpose of the Study:

• To enhance the speed and efficiency of the MGVB algorithm.
• To port the existing MGVB algorithm to a graphical processing unit (GPU).

Main Methods:

• The MGVB algorithm was recoded in CUDA C.
• Recursive functions and data structures were re-implemented non-recursively.
• The enhanced algorithm was integrated into a new version, cMGVB.

Main Results:

• cMGVB demonstrates significantly increased speed and efficiency compared to the original MGVB.
• A typical focused search is completed in approximately one minute by cMGVB, versus 10-15 minutes for the original implementation.
• cMGVB can run effectively on single workstations with affordable GPUs, outperforming the original algorithm on HPC clusters.

Conclusions:

• cMGVB enables computational proteomics workflows previously considered impractical or impossible.
• The GPU implementation offers substantial performance improvements for post-translational modification analysis.
• This advancement makes complex proteomic data analysis more accessible and efficient.