FPGA Accelerated Analysis of Boolean Gene Regulatory Networks

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Boolean models are essential for qualitative modeling of gene regulatory networks.
• Increasing model complexity due to high-throughput technologies challenges existing software simulation tools.
• Field Programmable Gate Arrays (FPGAs) offer potential for massive performance improvements in complex network simulations.

Purpose of the Study:

• To present a new simulation framework for Boolean models utilizing FPGAs.
• To address the scalability limitations of current software simulation tools for large Boolean models.
• To accelerate the analysis of complex molecular networks and attractor detection.

Main Methods:

• Developed a framework that converts Boolean models into Verilog, a hardware description language.
• Integrated the Verilog model with an execution core running on an FPGA.
• Coherently attached the FPGA to a POWER8 processor for simulation.

Main Results:

• Achieved an order of magnitude speedup compared to multi-threaded software simulations on a POWER8 processor.
• Demonstrated consistent performance improvements on a T-cell large granular lymphocyte leukemia (T-LGL) model, yielding new biological insights.
• Attained speedups of one to three orders of magnitude for attractor detection compared to software solutions.

Conclusions:

• The FPGA-based simulation framework significantly enhances the speed and scalability of Boolean model analysis.
• This approach enables faster discovery of biological insights from complex gene regulatory networks.
• The framework provides unprecedented speed for attractor detection, advancing systems biology research.