(Study of lattice proteins using D-Wave quantum annealers)

Quantum annealers, such as the 5000 Qbit D-Wave Advantage machine "JUPSI" in Jülich, are devices which can directly use quantum mechanical properties to improve certain kinds of computation. Using our recent studies of protein systems on JUPSI as demonstrative examples, I will present quantum annealing as a computational tool and discuss its strengths and weaknesses. I will present how we coded our intuitively simple protein model for the D-Wave quantum annealer, with the hope that some of our insights will be transferable to other research projects.

For the purpose of our experiments with the quantum annealer, we represented a protein as a self avoiding chain of beads placed on a 2D square lattice, where each bead is either hydrophobic (H) or polar (P). Hydrophobic beads on neighbouring lattice sites have an attractive interaction in addition to the repulsive "self-avoidance" interaction which prevents two beads from occupying the same lattice site. The protein is identified by the sequence of H/P beads along the chain, e. g., HPHHPHPHHPP. The potential energy of a structure is calculated by aggregating contributions from the whole system, with a unit negative contribution for every pair of H residues occupying nearest neighbour lattice sites. We examined two very important questions in this simplified representation:

(i) The folding problem which relates to the identification of the ground state of the energy function for a given protein sequence

(ii) The design problem, which is the inverse of the above, where one searches for protein sequences which would have a given state as their ground state.

We formulated the problems in a manner suitable for quantum annealers, and performed computations on JUPSI, and compared our results with those from classical Monte Carlo methods. The D-Wave hybrid annealer computations outperform all alternatives we have tried by finding the correct solutions with a 100 % success rate, for chain lengths up to 64 amino acids.

Our results have been published in the following articles:

"Folding lattice proteins with quantum annealing", Anders Irbck, Lucas Knuthson, Sandipan Mohanty, and Carsten Peterson, Phys. Rev. Research 4, 043013 (2022)

"Using quantum annealing to design lattice proteins", Anders Irbck, Lucas Knuthson, Sandipan Mohanty, and Carsten Peterson, Phys. Rev. Research 6, 013162 (2024)