Stefan Bekiranov, Associate Professor of Biochemistry and Molecular Genetics and his team of researchers at the University of Virginia School of Medicine had successfully developed and implemented a genetic sample classification algorithm on a quantum computer. Researchers believe although quantum computing is still at its infancy now, the study marks a cornerstone to highlight the potential of the technology to advance the field of future genetic research.
Quantum computers today
AIMed briefly explored quantum computing in our earlier blog entry. As mentioned, classical computers make use of bits or electrical/optical pulses represented by 1 and 0. The binary digits form practically everything we virtually come across, from emails, tweets, to songs and videos. Quantum computers, on the contrary, run on qubits or subatomic particles like electrons and photons.
Qubits represent many combinations of 1 and 0 at once. This capability to exist in multiple states simultaneously is known as superposition. Precision lasers or microwave beams are two ways which researchers could superposition qubits. Sometimes, researchers would also entwine pairs of qubits, making the duo exist in one quantum state. Known as entanglement, over here when the state of one qubit changes, it will promptly change the state of the other too, regardless if they may be separated by some distance.
These properties ascertain qubits’ processing power; tremendously cutting down the amount of time classical computers require to solve classes of problems. As of now, quantum computers are foreseen to be most useful in the simulation of matter behavior as well as optimization, or churning out the many possible solutions in really short period of time. However, we do not have the adequate science and engineering knowledge to generate and maintain qubits.
In order to keep qubits in a controlled quantum state for functioning, some companies (i.e., Google, IBM) choose to use superconducting circuits to keep them in extremely low temperatures. While others (i.e., IonQ) will suspend them in electromagnetic fields on a silicon chip in highly vacuum chambers. As such, we are still unsure if quantum computing is a blessing or a stress for medicine and healthcare.
A quantum classifier for genomic data
According to Dr. Bekiranov, Dr. Maria Schuld, Quantum Machine Learning Developer at Xanadu, a Toronto-based photonic quantum computing and advanced artificial intelligence (AI) startup was one of the first to generate “implementable, near-term” quantum machine learning algorithms. What the research team at the University of Virginia School of Medicine did was to build upon those algorithms to create a quantum classifier to be executed on an IBM quantum computer.
This new quantum classifier, meant to be used on genomic data, will determine whether a test sample derives from a disease or control group at a much faster speed than classical computers. For example, classical computers require 3 billion operations to categorize four building blocks of human DNA (i.e., A, G, C, T) whereas the quantum algorithm only needs 32.
Dr. Bekiranov thought the success was preliminary because they only realized there were many technical challenges and limitations after they started testing the classifier on an IBM quantum computer. All they could do was to administer a more simplified or what he termed as the “toy” version of what they originally intended to. “Relatively small-scale quantum computers that can solve toy problems are in existence now,” Dr. Bekiranov said in the press release.
“The challenges of developing a powerful universal quantum computer are immense. Along with steady progress, it will take multiple scientific breakthroughs. But time and again, experimental and theoretical physicists, working together, have risen to these challenges. If and when they develop a powerful universal quantum computer, I believe it will revolutionize computation and be regarded as one of the greatest scientific and engineering achievements of humankind.”
The study is available here.