LONDON – The same algorithms that determine your Netflix viewing habits could soon create a cancer treatment plan.
Scientists have created a machine learning tool to study cancer-triggered DNA changes that uses the artificial intelligence programs used by the streaming giant. The program categorizes DNA changes into a cell’s complete genetic code as tumors start and grow.
The international team identified 21 common defects related to structure, order and number of specimens present. They are called copy number signatures, offering the hope of personalized therapies for patients.
Netflix generates data on the type of movies and TV shows you like, how often you watch them, and whether you give them a “thumbs up” or a “thumbs down.” A mathematical formula analyzes huge amounts of information to find patterns in content, then makes recommendations as you scroll.
The cancer algorithm is similar, sifting through thousands of lines of genomic data and selecting common patterns. Identifying how segments of DNA, or chromosomes, fit together helps establish the types of defects that can occur. In tests, the scientists looked for patterns in the fully sequenced genomes of 9,873 patients with 33 different types of cancer.
Algorithm can predict cancer behavior
Findings in the journal Nature will create a model that researchers can use to assess how aggressive cancer is, find its weak points and design new treatments.
“Cancer is a complex disease, but we have demonstrated that there are remarkable similarities in the chromosome changes that occur when it starts and how it develops,” says co-lead author Professor Ludmil Alexandrov of the University of California-San Diego in one Press release.
“Just as Netflix can predict which shows you choose to watch next, we believe we’ll be able to predict how your cancer is likely to behave, based on the changes its genome has already undergone,” Alexandrov continues.
“We want to get to the point where doctors can look at a patient’s fully sequenced tumor and match key tumor features to our model of genomic defects. Armed with this information, we believe that doctors will be able to offer better and more personalized cancer treatment in the future.
Scientists have previously studied how these large-scale defects occur in sarcoma and wanted to find ways to study these changes in different types of cancer. Using software called SigProfilerExtractor, developed by Dr. Alexandrov, the algorithm uses complex calculations to analyze sequencing data from cancer patients.
It spots common patterns in the way chromosomes are rearranged in different types of the disease. Scientists further investigated which copy number signatures most strongly affected patient outcomes.
Of the 21 specific signatures, tumors where chromosomes broke and reformed, known as chromothripsis, were associated with the worst survival rates. For example, the study found that patients with glioblastoma, an aggressive type of brain tumor, had worse outcomes if their tumor had undergone chromothripsis. On average, patients with glioblastoma without chromothripsis survived six months longer than those without.
Scientists seek to create ‘a personalized cancer plan’
Scientists hope further improvements will allow doctors to discover how cancers are likely to behave based on the original genetic traits and those they detect as they spread.
“To stay ahead of cancer, we need to anticipate how it adapts and changes,” says co-lead author Professor Nischalan Pillay of University College London (UCL).
“Mutations are the main drivers of cancer, but much of our understanding focuses on changes to individual genes in cancer. We missed the big picture of how large swaths of genes can be copied, moved, or deleted without catastrophic tumor consequences.
“Understanding how these events occur will help us regain an edge over cancer. Thanks to advances in genome sequencing, we can now see these changes happening in different types of cancer and figure out how to respond to them effectively.
SigProfilerExtractor and other software tools have been made freely available to other scientists. They can use the algorithm to build their own Netflix-like libraries of chromosomal changes from DNA, based on data obtained from tumor sequencing.
“We believe that making these powerful computational tools free to other scientists will accelerate progress towards a personalized cancer plan for patients, giving them the best chance of survival,” concludes first author Dr Christopher Steele of UCL. .
South West News Service writer Mark Waghorn contributed to this report.