Researchers at the University of Virginia School of Medicine say they are tapping into the potential of quantum computers to help us understand genetic diseases.
Stefan Bekiranov, PhD, and colleagues have report the development of an algorithm in their new study, “ Implementation of a Hamming distance–like genomic quantum classifier using inner products on ibmqx2 and ibmq_16_melbourne ” published in Quantum Machine Intelligence, to allow researchers to study genetic diseases using quantum computers, once there are much more powerful quantum computers to run it. The algorithm, a complex set of operating instructions, will help advance quantum computing algorithm development and could advance the field of genetic research one day, according to Bekiranov.
Quantum computers are still in their infancy. But when they come into their own, possibly within a decade, they may offer computing power on a scale unimaginable using traditional computers.
“We developed and implemented a genetic sample classification algorithm that is fundamental to the field of machine learning on a quantum computer in a very natural way using the inherent strengths of quantum computers,” Bekiranov said. “This is certainly the first published quantum computer study funded by the National Institute of Mental Health and may be the first study using a so-called universal quantum computer funded by the National Institutes of Health.”
Traditional computer programs are built on 1s and 0s, either-or. But quantum computers take advantage of a fundamental of quantum physics: Something can be and not be at the same time. Rather than 1 or 0, the answer, from a quantum computer’s perspective, is both, simultaneously. That allows the computer to consider vastly more possibilities, all at once.
The challenge is that the technology is technically demanding. Many quantum computers have to be kept at near absolute zero, the equivalent of more than 450 degrees below zero on the Fahrenheit scale. Even then, the movement of molecules surrounding the quantum computing elements can disturb the calculations, so algorithms not only have to contain instructions for what to do, but for how to compensate when errors occur.
“Our goal was to develop a quantum classifier that we could implement on an actual IBM quantum computer. But the major quantum machine learning papers in the field were highly theoretical and required hardware that didn’t exist. We finally found papers from Maria Schuld, PhD, who is a pioneer in developing implementable, near-term, quantum machine learning algorithms. Our classifier builds on those developed by Schuld,” Bekiranov said. “Once we started testing the classifier on the IBM system, we quickly discovered its current limitations and could only implement a vastly oversimplified, or ‘toy,’ problem successfully, for now.”
The new algorithm essentially classifies genomic data. It can determine if a test sample comes from a disease or control sample exponentially faster than a conventional computer. For example, if they used all four building blocks of DNA for the classification, a conventional computer would execute 3 billion operations to classify the sample. The new quantum algorithm would need only 32.[…]