Four lines of work in quantum computing for genomics, with the publications that describe them.
01
Pangenome-guided sequence assembly
Given short reads from an individual and a pangenome graph, the task is to find the walk through the graph that best explains the reads. We reformulate this as quadratic and higher-order unconstrained binary optimisation (QUBO and HUBO) and solve it with iterative QAOA, with classical pre-processing and post-processing for problem reduction, read mapping and solution refinement.
Compared with de novo assemblers, the method returns far fewer contigs, which simplifies downstream analysis, and includes a methodology for reintegrating the computational solutions back into the biology.
Building a phylogenetic tree under the maximum compatibility criterion reduces to finding maximum weighted cliques in a graph, a problem for which exhaustive classical search quickly becomes infeasible. We combine classical and quantum pre-processing that shrinks the input graph without loss with a QAOA solver and post-processing that returns only maximal, feasible cliques.
The solver uses one qubit per vertex and at worst quadratic circuit depth. It has been demonstrated on IBM Heron r3 hardware, where it recovered more maximum-weight cliques than classical solvers.
Publication in preparation.
03
MPS-based genome encoding
Each base of a genome is mapped to one of the four states of two qubits and the whole sequence is prepared as a single quantum state using matrix product state methods. Trading circuit depth for width keeps each segment within reach of current hardware, and index-reported verification confirms the prepared state.
With this method, and in collaboration with the University of Melbourne, the complete hepatitis delta virus genome, 1,614 bases, was loaded onto NISQ hardware, the first time a whole genome has been loaded onto a quantum computer. Genome states prepared this way can be aligned with quantum sequence alignment, so the encoding is ready for downstream use.
The genome loading result: publication in preparation.
04
ML-based DNA encoding
A quantum encoding of DNA sequences, built on the rotary positional encodings used in large language models, in which the fidelity between two encoded sequences tracks their edit distance. It has no known classical or quantum analogue, and sequences of billions of bases can be represented on a few dozen qubits.
Used classically, the encoding gives RotorMap, a GPU-accelerated mapper reported to be 50 to 700 times faster than single-threaded Minimap2 on long, noisy reads. On quantum hardware, the angular form of the encoding produces state preparation circuits directly and has been tested on Quantinuum H2 and Helios systems. The same encoding gives a quantum DNA authentication protocol with evidence of a quantum advantage in communication complexity.
The research behind this work is published openly. Every entry below links to a version that is free to read
online: preprints on arXiv, and a journal article published under an open licence.