FPGA accelerator for protein secondary structure prediction based on the GOR algorithm.

Area of Science:

• Bioinformatics
• Computational Biology
• Biophysics

Background:

• Protein structure dictates function, making prediction crucial.
• The GOR algorithm is a key tool for protein secondary structure prediction.
• Increasing protein data necessitates faster prediction methods.

Purpose of the Study:

• To accelerate the GOR-IV algorithm for 2D protein structure prediction.
• To develop a fine-grained parallel hardware implementation on FPGA.
• To address the computational demands of large protein databases.

Main Methods:

• Implemented a parallel hardware architecture on an FPGA (XC5VLX330).
• Partitioned parameter tables for parallel access and exploited data reuse.
• Pipelined computation to overlap sequence loading, computing, and back-writing.

Main Results:

• Achieved a speedup of over 430x compared to the original GOR-IV.
• Demonstrated a 110x speedup versus a multi-thread SIMD PC implementation.
• Reduced power consumption to approximately 30% of general-purpose CPUs.

Conclusions:

• FPGA acceleration offers significant performance gains for GOR-IV.
• The proposed method provides a highly efficient and power-saving solution for protein structure prediction.
• This approach is well-suited for accelerating bioinformatics algorithms.