Earlier this year, we projected that 2021 would be the year of quantum computing for drug discovery, positing the idea that this is one of only a few areas where there is big ROI to offset the work required to map codes to quantum systems.
We’ve pointed to a number of companies that are exploring quantum for drug discovery, although it seems much of the momentum on the systems side has been around IBM. The tide might be shifting with quantum annealing systems maker D-Wave finding less publicized inroads at companies, including GlaxoSmithKline (GSK) and outperforming its quantum rival.
Drug maker GSK is evaluating quantum computing for specific workloads that hit scalability and complexity walls with traditional architectures and approaches, specifically genetic algorithms. While D-Wave showed the most promising results, we should note early that they only used IBM’s quantum simulation environment to evaluate the gate model approach.
Specifically, the company wanted to understand how quantum computing might give more comprehensive results with both quantum annealing (D-Wave) and standard gate model systems (IBM’s quantum simulator in this case).
The mRNA codon optimization problem is a good fit for quantum annealing because it is an NP-hard optimization problem—the very problem set D-Wave has often discussed when describing potential use cases. In this problem each amino acid in a protein sequence can be represented by as many as six codons, the goal is to find the right combination. This problem is important because, as the authors of the GSK paper on codon optimization on quantum architectures note, it can “impact downstream processes such as protein folding and function, which is important for use cases including recombinant protein therapies.”
“While classical approaches such as genetic algorithms can be highly performant, the fraction of solution space that is sample in a fixed number of iterations decreases exponentially as the polypeptide chain length grows. Thorough sampling of the solutions space is therefore often intractable with biologically relevant use cases,” the team explains.
While the nature of the problem and how it was mapped to both the D-Wave and IBM systems is detailed in the paper, overall they found the D-Wave approach to be “competitive in identifying optimal solutions and future generations of AQCs may be able to outperform classical approaches.”
IBM’s gate model approach wasn’t fully tested on their cloud-based systems. Instead GSK researchers used the Qasm noisy simulator, which can simulate up to 24 fully connected qubits. “While current generations of devices lack the hardware requirements in terms of both qubit count and connectivity to solve realistic problems, future generation devices may be highly efficient,” they conclude.
“Implementing this program on the D-Wave Advantage shows that it identifies high quality solutions that are competitive with a basic implementation of a genetic algorithm programmed with an equivalent scoring function,” the GSK team says. They add that “Implementing a version of this program for IBM Q devices, while successful, shows that modelling practical systems requires too many qubits to be run on even the most advanced gatebased devices available (e.g. IBM’s 65-qubit Hummingbird device). However, executing the model on an IBM noisy simulator41 demonstrates the potential of the algorithm.”
GSK has been less bold with its proclamations of quantum ambitions, at least compared to other major drugmakers striking big visibility partnerships with IBM, Google, and others. Nonetheless, seeing quantum computing in drug discovery at a GSK-sized company and noting that the challenge is no longer accessing the hardware on-prem and is more a problem of carefully mapping the work is promising—not just for GSK but for quantum overall.
Originally published by
Nicole Hemsoth | February 24, 2021
The Next Platform
