GPrimer

GPrimer accelerates primer design for quantitative PCR (qPCR) by leveraging GPU acceleration to generate validated primer pairs from whole sequence databases and perform homology testing on off-target sequences.

Key Features:

• GPU acceleration: Uses GPUs to accelerate primer-design computations, achieving a 57-fold overall speedup and a 557-fold speedup in the most time-intensive step versus MRPrimer running on a six-machine cluster (human RefSeq benchmark).
• MRPrimer MapReduce foundation: Builds on the MRPrimer MapReduce-based pipeline and retains MRPrimer's homology-testing approach on off-target sequences.
• Whole-database processing: Accepts an entire sequence database as input and outputs all feasible and valid primer pairs without requiring BLAST-like additional steps.
• Optimized MRPrimer steps: Optimizes two time-consuming steps within the MRPrimer framework to improve throughput.
• Coalesced memory access: Implements data structures that enable coalesced memory access on GPUs.
• Workload balancing: Performs efficient workload balancing among GPU threads.
• Streaming transfers: Uses streaming data transfers between main memory and GPU memory.

Scientific Applications:

• qPCR primer design: Design of validated primer pairs for quantitative PCR experiments.
• Gene expression studies: Rapid primer generation to support gene expression analysis using RefSeq and other sequence databases.
• Genetic screening and molecular diagnostics: High-throughput primer design for genetic screening and molecular diagnostic assays.

Methodology:

Built on the MRPrimer MapReduce framework, GPrimer optimizes two MRPrimer steps using GPU acceleration with data structures for coalesced memory access, efficient workload balancing among GPU threads, streaming transfers between main memory and GPU memory, and retains homology testing on off-target sequences while outputting all feasible primer pairs from whole sequence databases.