A hardware demonstration of a universal programmable RRAM-based probabilistic computer for molecular docking.

Area of Science:

• Computational biology
• Biophysics
• Computer engineering

Background:

• Molecular docking is crucial for drug discovery but computationally intensive due to vast search spaces.
• Existing computing methods face limitations, and quantum computing is not yet scalable or precise enough.
• Probabilistic computing offers a potential alternative for complex computational biology problems.

Purpose of the Study:

• To develop and demonstrate a probabilistic computer (p-computer) prototype for solving complex molecular docking problems.
• To overcome the limitations of conventional and quantum computing in computational biology.
• To showcase a novel hardware application of p-computing in bioinformatics.

Main Methods:

• Fabrication of a p-computer prototype using 180nm CMOS with HfO₂ RRAM and compute-in-memory (CIM) schemes.
• Integration of Gaussian Random Number Generator-based p-bits with CIM, encoding coefficients in RRAM crossbars.
• Experimental solution of a 42-node molecular docking problem involving the lipoprotein LolA-LolCDE complex.

Main Results:

• The p-computer prototype successfully solved a complex 42-node molecular docking problem.
• Results were consistent with established tools like the Protein-Ligand Interaction Profiler.
• The co-designed CIM and p-bit architecture alleviated memory-to-compute bottlenecks.

Conclusions:

• The developed p-computer demonstrates a viable hardware application for complex bioinformatics problems.
• This approach shows potential to overcome success rate and efficiency limitations of current docking technologies.
• The study highlights the promise of p-computing for advancing drug discovery and computational biology.