For a child diagnosed with neuroblastoma—the most common infant cancer, occurring when early nerve cells grow out of control—the path to treatment isn’t simple. Some types of neuroblastoma resolve on their own, while others require aggressive intervention. Researchers have tried matching treatments to patients based on one-gene mutations with limited success. This is because patients’ outcomes depend on their entire molecular background containing millions or even billions of features, such as DNA and RNA from tissues and blood.
“It’s much more than just one gene—everything that’s happening in the cells of the patient matters,” said Orly Alter, an associate professor of biomedical engineering at the University of Utah’s Scientific Computing & Imaging Institute and a member of the Huntsman Cancer Institute‘s Cancer Control & Population Sciences.
Current artificial intelligence and machine learning (AI/ML) approaches require massive amounts of training data, and, specifically, vastly more patient samples than genetic features. This makes them poorly suited for predicting patient outcomes in most clinical trials, which typically enroll just 20 to 100 people. For example, a recent large language model of the 30,000-nucleotide genome of the COVID-19 virus required about 110 million samples. Translating this to the 3-billion-nucleotide human genome, a conventional AI approach would need 33 trillion patients.
By using the mathematics of quantum mechanics, Alter and her collaborators developed a novel AI/ML technique that can improve treatment selections and drug success rates. Their work appears in the journal Applied Physics Letters (APL) Quantum.
Billions of molecular features
“Our quantum approach allows us to find the relevant information in every layer of the data, for example, from the patients’ blood in addition to their tumors,” Alter said. “Even for very few patients, we can still take everything in—their millions to billions of molecular features—and make sense of them. We can, therefore, understand the disease mechanisms and predict drug targets to improve patients’ outcomes. We also validate our AI/ML predictions of targets and outcomes experimentally, which is widely considered a biotechnology holy grail.”
The technique deploys a set of algorithms, called multitensor comparative spectral decompositions, which Alter built on the quantum mechanical concepts of entanglement and superposition. Like a prism splitting white light into individual colors, this approach breaks down a patient’s multiple layers of molecular data—such as their tumor and blood genomes and tumor transcriptome (or the RNA messages driving the cancer’s growth)—into linked patterns that predict health outcomes.
Alter and her team demonstrated their technique with an analysis of open-source data of neuroblastoma cases. The algorithms discovered two new predictors of patients’ life expectancy in response to treatment, and these predictors consistently outperformed standard biomarkers across tumor and blood DNA and tumor RNA. These findings held up across separate groups of children treated at different times and hospitals, meaning that the method can be applied to the general population in order to provide a clearer roadmap for patient care and drug development.
Continue reading at the Scientific Computing & Imaging Institute.