GPrimer: a fast GPU-based pipeline for primer design for qPCR experiments.

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• High-quality primer design is critical for quantitative polymerase chain reaction (qPCR) experiments.
• The MRPrimer pipeline, while effective, suffers from slow computational speed when handling large sequence databases.
• Exponential growth in sequence data necessitates faster primer design solutions.

Purpose of the Study:

• To develop a significantly faster primer design pipeline than MRPrimer.
• To leverage GPU (Graphics Processing Unit) computational power for accelerated primer design.
• To maintain the same input and output as MRPrimer for seamless integration.

Main Methods:

• Developed GPrimer, a GPU-based pipeline for primer design.
• Optimized computationally intensive steps of MRPrimer using GPU acceleration.
• Implemented data structures for coalesced memory access and workload balancing on GPUs.
• Utilized streaming fashion for data transfer between main memory and GPU memory.

Main Results:

• GPrimer achieves substantial speedup compared to MRPrimer.
• A 57-fold speedup for the entire pipeline and a 557-fold speedup for the most time-consuming step were observed on a single machine with 4 GPUs.
• These improvements were demonstrated using the human RefSeq database.

Conclusions:

• GPrimer provides a GPU-accelerated solution for primer design.
• The pipeline efficiently identifies all feasible and valid primer pairs from a sequence database in a single step.
• The GPrimer software is publicly available for use.