The science behind predicting your viewing habits on Netflix could one day be used to guide doctors in managing some of the hardest-to-treat cancers, shows a study led by the University of California San Diego and University College London.
Image credit: Fernando Zhiminaicela via Pixabay, free licence
The researchers used artificial intelligence to analyze and categorize the size and scale of DNA changes across the genome when cancer starts and grows. By analyzing genomes from 9,873 patients with 33 types of cancer, the scientists found 21 categories of common changes to the structure and number of chromosomes in the genetic material of tumors.
These common DNA changes, known as copy number signatures, could be used to build a blueprint to predict how a cancer is likely to progress and design the most effective treatments for it. The findings are reported in a paper published in Nature.
“Cancer is a complex disease, but we’ve demonstrated that there are remarkable similarities in the changes that happen in chromosomes when different cancers start and grow,” said Ludmil Alexandrov, a professor of bioengineering and cellular and molecular medicine at UC San Diego. The latter is a co-lead author of the study.
When cancer starts, mutations in the DNA can cause large-scale faults to occur across the whole genome. These faults can result in too few or too many chromosomes compared to normal cells. Tumors can also develop faults in the mechanisms designed to repair their DNA, leading to further faults in the structure of DNA within chromosomes, as well as errors when the DNA tries to make copies of itself.
The researchers were interested in studying these large-scale genomic faults across different types of cancer. Enter a suite of AI tools developed by Alexandrov’s lab, called SigProfiler, which scans sequencing data from cancer patients and identifies common patterns in chromosome changes in different types of cancer.
“Based on these changes that the genome has previously experienced, our algorithm can predict how your cancer is likely to behave—similar to how Netflix can predict which series you’ll choose to binge watch next based on your previous viewing activities,” said Alexandrov.
This algorithm was key in identifying the 21 copy number signatures found in this study. This also enabled the researchers to predict how some of the hardest-to-treat cancers will behave.
One the copy number signatures created by the algorithm is attributed to an event known as chromothripsis, where chromosomes in tumors fragment and rearrange. This copy number signature was associated with the worst survival outcomes, the researchers found. Take patients with a lethal, fast-growing brain cancer called glioblastoma, for example. On average, glioblastoma patients whose tumors did not undergo chromothripsis were found survive six months longer than those whose tumors did.
“Mutations are the key drivers of cancer, but a lot of our understanding is focused on changes to individual genes in cancer. We’ve been missing the bigger picture of how vast swathes of genes can be copied, moved around or deleted without catastrophic consequences for the tumor,” said Nischalan Pillay, a professor of sarcoma and genomics at University College London and co-lead author of the study. “Understanding how these large-scale genomic events arise will help us regain an advantage over cancer.”
The researchers have made SigProfiler and other software tools used in the study feely available to other scientists so that they use the tools to build their own Netflix-style libraries of chromosome changes in DNA based on data obtained from sequencing tumors.
“As it becomes faster and cheaper to read an individual’s genetic code in full, we hope our blueprint will be widely used to navigate that code and help doctors offer better and more personalized cancer treatment,” said Alexandrov.
As part of their next steps, the researchers explore some of the identified categories of copy-number changes as clinical biomarkers for predicting response to anti-cancer therapies.
Source: UCSD