Approach
We build unified quantum and classical pipelines that reframe hard genomics problems as optimisation and encoding tasks, and we validate them on real quantum hardware.
-
01
Biological task
Assemble DNA reads, align sequences or construct a phylogenetic tree.
-
02
Quantum data encoding
Lossless MPS-based encoding, compressed ML-based encoding or a graph representation.
-
03
Quantum algorithms
Quantum optimisation and quantum sequence alignment, with classical pre- and post-processing.
-
04
Biological solutions
Assembled genomes, aligned genomic data and phylogenetic trees.
Where quantum computing fits
Most genomic analysis is routine and well served by classical software. The difficulties lie in a small number of problems whose cost grows combinatorially: aligning many sequences at once, resolving complex and highly variable regions such as the HLA genes, and matching reads against a pangenome rather than a single reference. These tasks demand a great deal of computation and relatively little storage, which is the profile that suits quantum computers. We concentrate on those problems and leave the rest to classical tools.
Encoding genomic data
Getting genomic data onto a quantum computer is a problem in its own right, and we use two encodings. The first is exact and lossless. Each base 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 so that each segment stays within reach of current hardware.
The second encoding is compact, and preserves similarity between sequences rather than every base. Derived from the rotary positional encodings used in large language models, it maps DNA sequences to quantum states whose fidelity tracks their edit distance, so that sequences of billions of bases can be represented on a few dozen qubits. Used classically, the same encoding gives RotorMap, a fast mapping tool for long, noisy reads.
Optimisation
Many tasks in genomics can be posed as optimisation problems: choosing the best assembly, the best tree or the best matching from a very large number of candidates. We write these problems in forms that quantum algorithms can act on and solve them with hybrid methods, in which classical pre-processing reduces the problem before the quantum step and classical post-processing turns the output into a valid answer. Where a problem has many good solutions rather than one, sampling from a quantum device returns a range of candidates to choose from.
Validation on hardware
Every method is run on today's devices as well as on simulators: gate-based machines from IBM, Quantinuum and IonQ, and D-Wave annealers, at scales beyond 100 qubits. We report accuracy, resource counts and run time, and we compare against established classical tools. The complete hepatitis delta virus genome was loaded onto NISQ hardware in collaboration with the University of Melbourne (publication is in preparation).
Back into biology
Computational solutions are reintegrated into the biological analysis as assembled genomes, aligned data and phylogenetic trees. A web pipeline lets bioinformaticians upload data, choose a solver and run these methods without prior knowledge of quantum computing, with jobs executed on high-performance computing resources and on quantum hardware in the cloud.