IBM Watson Takes on the Genetics of Brain Cancer

The AI will analyze patients' genetic data to provide treatment suggestions

3 min read
IBM Watson Takes on the Genetics of Brain Cancer
Photo: IBM

Twenty patients with an aggressive form of brain cancer will have a new doctor on their medical team: the learned geneticist known as IBM Watson. In a collaboration announced today between IBM and the New York Genome Center, IBM's Jeopardy-beating AI will analyze the genomes of those 20 patients in hopes of providing insights for their oncologists. 

IBM has been promoting its AI as a killer app for health care, thanks to Watson's natural language processing skills and machine learning abilities. Over the past two years Watson has been engaged in a separate project at New York's Memorial Sloan-Kettering Cancer Center, in which doctors are training the AI to understand the language of medicine. In that project, Watson is being taught to read patients' records and search the medical literature for relevant suggestions on treatment. 

This new project will show that Watson can provide deeper analysis for such point-of-care applications, said IBM Research's Raminderpal Singh after a press conference in New York today. For these 20 cancer patients, Watson won't just scan the medical literature for information. The AI will also scan the genetic data from the patients' own healthy cells and cancer cells, and will then search for information that's relevant to the genetic mutations in the tumor. "As genome sequencing becomes more commonplace—and it will—we'll need a way to go from mutation information to clinically actionable information," said Singh.   

Robert Darnell, president of the New York Genome Center, explained that the project will take the form of a research study spanning nine hospitals and institutions, and involving 20 patients with glioblastoma, a malignant brain cancer that kills more than 13,000 people in the United States each year. Glioblastoma also kills quickly, Darnell said, with a typical life expectancy after diagnosis of just 12 to 14 months.

Sequencing the DNA of a patient's tumor provides "a gusher of information," Darnell said, as there may be thousands of mutations that cause a glioblastoma tumor cell to differ from a healthy cell. What's more, each patient's tumor will have its own unique set of mutations. Oncologists are still learning how to analyze and make use of that data, Darnell said. "We need time to think about it—and time is not your friend when it comes to glioblastoma." A dedicated team of geneticists might take several weeks to come up with a treatment plan for a patient, he said, which is enormously commendable, but not very scalable. 

Watson can take a patient's set of mutations and scan thousands of medical papers in seconds to find out how those mutations affect the cancer cell, and which treatments have been most effective on genetically similar tumors. The AI then presents the clinician with a summary of the patient's mutations and various treatment options. In the image below, a cancer cell's mutations are mapped on a cell protein pathway. 

The current research project is intended to evaluate Watson's usefulness as a tool for oncologists; the final treatment decisions still rest with the human doctor. Presumably that will be the case for the foreseeable future, as patients probably aren't ready to put themselves entirely in the hands of a robo-doc. What's more, it's fairly easy to envision scenarios in which human intelligence would trump the AI, which operates best when dealing with lots of data. For example, Watson might find hundreds of studies regarding an older drug for glioblastoma, and only a single study regarding a revolutionary new treatment that trumps the old drug. A human doctor who is up on the literature might recommend the new treatment right away, whereas Watson wouldn't be confident in doing so until a critical mass of follow-on papers had deemed it effective. 

But if the project is successful and Watson continues to ingest data, the AI will only get smarter over time. IBM's Singh says the team is already building in a function to include knowledge of past outcomes in Watson's decision-making process. "Watson will know whether clinicians used its findings, and also the patients' outcomes," he says. 

Images: IBM, Jon Simon/Feature Photo Service for IBM

The Conversation (0)

Understanding the Coronavirus Is Like Reading a Sentence

And parsing its "words" and "grammar" could lead to better COVID-19 vaccines

10 min read
Illustration showing the structure of the SARS-CoV-2 virus particle. At the virus's core is its RNA (ribonucleic acid) genome (coils). Embedded in the viral envelope (grey) are spike proteins (red) that the virus uses to attach to and infect a host cell.
John Bavaro/Science Source

Since the beginning of 2020, we've heard an awful lot about RNA. First, an RNA coronavirus created a global pandemic and brought the world to a halt. Scientists were quick to sequence the novel coronavirus's genetic code, revealing it to be a single strand of RNA that is folded and twisted inside the virus's lipid envelope. Then, RNA vaccines set the world back in motion. The first two COVID-19 vaccines to be widely approved for emergency use, those from Pfizer-BioNTech and Moderna, contained snippets of coronavirus RNA that taught people's bodies how to mount a defense against the virus.

But there's much more we need to know about RNA. RNA is most typically single-stranded, which means it is inherently less stable than DNA, the double-stranded molecule that encodes the human genome, and it's more prone to mutations. We've seen how the coronavirus mutates and gives rise to dangerous new variants. We must therefore be ready with new vaccines and booster shots that are precisely tailored to the new threats. And we need RNA vaccines that are more stable and robust and don't require extremely low temperatures for transport and storage.

That's why it's never been more important to understand RNA's intricate structure and to master the ability to design sequences of RNA that serve our purposes. Traditionally, scientists have used techniques from computational biology to tease apart RNA's structure. But that's not the only way, or even the best way, to do it. Work at my group at Baidu Research USA and Oregon State University has shown that applying algorithms originally developed for natural language processing (NLP)—which helps computers parse human language—can vastly speed up predictions of RNA folding and the design of RNA sequences for vaccines.

Keep Reading ↓ Show less