Watson , the artificial intelligence system that became famous when it beat humans on the game show Jeopardy!, is now taking on the task of analyzing oceans of genetic data and suggesting treatments for cancer—a venture that could usher in a new era of personalized medicine, if it succeeds .
The announcement was made Thursday at the offices of the New York Genome Center, a nonprofit research organization partnering with IBM where Watson was developed.
John Kelly described Watson as a "cognitive system" that mimics some of the capabilities of the human brain: taking in large amounts of information, recognizing relevant data, using it, and learning from its use.
Tailored treatments for patients
In the new program, Watson will offer suggestions for personalized treatments, which are currently rarely available to patients, and can require a week of work from a team of genetic counselors and other specialists.
The program will begin with 20 to 25 patients with glioblastoma, a form of brain cancer that is usually fatal. The average survival time after diagnosis is just 14 months, said Robert Dannell , the institute's CEO
Unlike other research efforts, which focus on discovering new mutations that trigger cancer, Watson will only try to find the optimal personalized treatment.
The system will be fed DNA and mRNA genetic sequences from patients' tumors and healthy tissues. In many patients, tumors actually contain populations of different cells, which may have different combinations of mutations—a factor that is not usually considered in treatment planning, but will be taken into account by Watson.
In addition, Watson has been trained in biochemistry, taking advantage of the fact that the US National Institutes of Health (NIH) has compiled tables of human biochemical pathways in a machine-readable format. Armed with this knowledge, the system can read the NIH's PubMed database, which contains abstracts and full texts of almost all biomedical publications.
High-intelligence electronic filing
By combining the literature with each patient's genetic data, Watson will attempt to identify the mutations that characterize cancer cell populations, examine the metabolic pathways that are affected, and identify drugs that target proteins in these pathways.
He will thus be able to offer advice on the optimal therapeutic approach, although the final decision will be made by doctors.
If the system produces positive results in the first 20-25 patients, the program could expand - all major hospitals in New York City have agreed to participate in the effort.
IBM has moved Watson to the cloud (it previously occupied a large computer cluster), which makes it easier for other institutions to use it.
