Netflix algorithm could help doctors treat cancer

Based on new study co-led by UCL scientists and supported by Cancer study UK and Cancer Grand Challenges, the science underlying your Netflix viewing habits may soon help clinicians manage cancer.

An international team of scientists employed artificial intelligence (AI) in the study, which was published in Nature, to look into and classify the magnitude and scope of DNA alterations throughout the genome, or a cell's whole genetic code, during the initiation and progression of cancer.

Scientists have used artificial intelligence (AI) to find 21 frequent errors in the DNA's structure, order, and quantity that arise when cancer begins and spreads. These widespread errors, known as copy number signatures, may be able to direct medical professionals toward tumor-specific therapy.

Data regarding the kinds of movies and TV shows you view, how often you watch them, and whether you give them a "thumbs up" or "thumbs down" are created when you watch Netflix. When you scroll through Netflix, an algorithm is used by the service to analyze the vast amount of data it has collected, look for trends in the material you view, and suggest new movies and TV shows.

A similar algorithm was developed by a group of researchers led by Drs. Ludmil Alexandrov (University of California, San Diego) and Nischalan Pillay (UCL Cancer Institute). This algorithm can sort through thousands of lines of genomic data and identify common patterns in the way the chromosomes organize and arrange themselves. After then, the system can classify the patterns that show up, assisting researchers in identifying the many kinds of errors* that may arise in cancer.

The researchers used the program to search through 9,873 individuals' completely sequenced genomes, which included 33 distinct cancer types, for trends. The system found 21 frequent errors in the chromosomal makeup and organization of tumors, classifying them into distinct “genres” known as copy number signatures.

Now, using the 21 copy number signatures, a blueprint will be created that researchers may use to determine the cancer's potential for aggressiveness, identify its weak points, and develop novel treatment strategies.

Co-lead author of the study Dr. Ludmil Alexandrov, an associate professor at UC San Diego, stated: "We've shown that there are remarkable similarities in the changes to chromosomes that happen when it starts and how it grows. Cancer is a complex disease."

"We think we will be able to forecast how your cancer is going to behave, based on the modifications its genome has already experienced, just as Netflix can predict which shows you'll want to binge watch next.

Our goal is for physicians to be able to examine a patient's completely sequenced tumor and compare its salient characteristics with our genetic flaw database. We think that with such knowledge, medical professionals will be able to provide more individualized and superior cancer therapy in the future.

The goal of the research was to determine how these extensive genetic errors manifest in various cancer forms, having previously examined how they do so in sarcoma.

The method employs sophisticated arithmetic to examine sequencing data from cancer patients and find similar patterns in the chromosomal reorganization in various cancer types. Dr. Alexandrov developed the software known as SigProfilerExtractor.

The copy number signatures that had the biggest impact on cancer patients' outcomes were the subject of additional research by the experts. The scientists discovered that, out of the 21 characteristics detected by the algorithm, the tumors with chromothripsis—a condition in which chromosomes break and reassemble—were linked to the worst survival rates. For instance, the study discovered that chromothripsis had a negative impact on survival rates for individuals with glioblastoma, an aggressive kind of brain tumor. Patients with glioblastoma whose tumours lacked chromothripsis often lived six months longer than those whose tumours did.

By fine-tuning the algorithm, the scientists hope to help physicians determine the likely course of your cancer based on the genetic characteristics it began with and the genetic alterations it picks up along the way.

Dr. Nischalan Pillay, an associate professor of sarcoma and genomics at UCL and co-lead author of the study, stated: "We need to anticipate how cancer adapts and changes in order to stay one step ahead of it."

"Although mutations are the primary causes of cancer, alterations to specific genes in cancer account for a large portion of our knowledge." The fact that large portions of genes may be duplicated, rearranged, or removed without having disastrous effects on the tumor has escaped our notice.

Gaining insight into the causes of these occurrences will provide us a fighting chance against cancer. Genome sequencing advancements have made it possible for us to see these alterations in a variety of cancer types and determine the most effective ways to treat them.

The researchers have released SigProfilerExtractor and other software tools used in the work under open license, enabling other researchers to utilize the method to create their own libraries of chromosomal alterations from DNA, a la Netflix, using information from sequencing tumors.

"We believe that making these powerful computing tools free to other scientists will accelerate progress towards a personalised cancer blueprint for patients, giving them the best chances of survival," stated Dr. Christopher Steele, the first author, from the UCL Cancer Institute.

According to Cancer Research UK CEO Michelle Mitchell, the organization has led the way in our understanding of the genetics underlying cancer. As we commemorate our 20th anniversary, it's astounding to learn that we have the ability to create a blueprint for a variety of tumor types, which may aid scientists in forecasting the behavior of cancer and illuminating potential avenues for targeted treatment.

"This study is yet another outstanding outcome of the Cancer Grand Challenges, an initiative designed to address some of the most significant obstacles in cancer research."