An FPGA-based hardware accelerator supporting sensitive sequence homology filtering with profile hidden Markov models

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Sequence alignment is crucial for genome annotation.
• Profile hidden Markov models (pHMMs) offer greater sensitivity than BLAST for sequence alignment.
• The HMMER tool's sequence-specific value (SSV) calculation is a computational bottleneck.

Purpose of the Study:

• To develop a hardware accelerator for the SSV calculation step in HMMER.
• To improve the speed and efficiency of pHMM-based sequence alignment.
• To demonstrate the utility of FPGA hardware accelerators in bioinformatics.

Main Methods:

• Designed and implemented HAVAC, a Field-Programmable Gate Array (FPGA) hardware accelerator.
• Targeted the sequence-specific value (SSV) matrix computation, a key bottleneck in HMMER.
• Evaluated HAVAC's performance on a Xilinx Alveo U50 FPGA card.

Main Results:

• HAVAC achieved 1739 GCUPS, 227x faster than nhmmer's SSV implementation.
• Considering data transfer, HAVAC was 65x faster (1 thread) and 35x faster (4 threads) than nhmmer.
• HAVAC utilized 31% of the energy consumed by a high-end Intel CPU.

Conclusions:

• FPGA hardware accelerators like HAVAC offer substantial speed improvements for sequence annotation.
• HAVAC frees the host CPU for simultaneous computation of other analysis pipeline aspects.
• This work highlights the potential of hardware acceleration in advancing bioinformatics applications.