Hybrid quantum-classical solvers help design proteins
A new study shows that hybrid quantum-classical solvers can find low-energy protein structures more effectively than traditional methods.
Researchers have found that hybrid quantum-classical solvers can successfully tackle a difficult task in protein design. This method works well for finding the best shapes for protein parts. It performs better than some standard tools on larger tasks.
Proteins are long chains made of amino acids. Each amino acid has a backbone and a side chain. The side chains determine how a protein behaves and its physical properties. Scientists want to change these side chains to create proteins with specific traits.
One major step in this process is protein side-chain optimization (PSCO). This involves choosing the best shapes for side chains to reach a low energy state. Because there are so many possible combinations, finding the best one is very hard.
The researchers looked at a specific type of math problem called a binary quadratic program (BQP). These problems involve binary variables and quadratic equations. Previous tests showed that hybrid solvers are good at BQPs, but few real-world problems fit this exact math structure.
The team proposed that PSCO is a perfect real-world example of a BQP. It uses binary variables to choose between different shapes. It also requires finding many different low-energy options rather than just one single best answer.
To test this, the researchers used 40 different protein instances. These were based on 20 known protein structures. The proteins had up to about 200 amino acids and 2000 binary variables.
They compared a hybrid solver running for 5 seconds against a classical solver called Gurobi. Gurobi was tested at 10 seconds, 1000 seconds, and 4 hours. The hybrid solver reached the best possible answer whenever one was known.
On smaller protein examples, the hybrid solver matched the performance of Gurobi. However, on larger examples, the hybrid solver was better. It found solutions with even lower energy than the classical solver could find.
A rotamer is a preferred combination of angles for a side chain. A side chain is the part of an amino acid that differs from others. Quantum annealing is a way of computing that uses physical properties to solve optimization problems.
The study shows that the advantage of hybrid solvers extends to real-world biological problems. This result suggests that more research is needed. The next step is to compare these results against other specialized classical methods.
The hybrid system uses both quantum and classical parts. It uses classical algorithms and only uses the quantum processing unit (QPU) when it helps. This makes the technology more practical for industrial use.
Key facts
- Hybrid quantum-classical solvers outperformed Gurobi on large protein side-chain optimization tasks.
- The study tested 40 protein instances with up to 2000 binary variables.
- Protein side-chain optimization is a natural fit for binary quadratic programs.
- The hybrid solver reached the certified optimum in all available cases during the test.