I have worked with various flavors of Watson in the last 4 years, and it has taught me a lot about the need to understand bias, both in data and the interpretation of it. In the spirit of transparency, I will be honest about my own bias – my favorite Watson Health product is Watson for Drug Discovery (WDD). Read on to find out why, and tune in to the Discovery Channel at 9pm EST on June 21st to see for yourself.
From ugly duckling to beautiful swan
WDD was an ugly baby in 2014 when we worked with Baylor and demonstrated that we could predict kinases of P53. We had some powerful algorithms that could detect signals in biomedical literature. However, algorithms are really just math and statistics, and let’s be honest, most of us do not want to experience math and statistics first hand in our daily jobs. We had a great product concept; the challenge was to make it consumable.
How do you take AI and insert it into a workflow to impact decision making?
We operate under 3 guiding principles:
1) PURPOSE: AI and cognitive systems are developed and applied by IBM to augment and not replace human intelligence.
2) TRANSPARENCY: For cognitive systems to fulfill their potential in changing the world, it is vital that people have confidence in their recommendations, judgments and uses.
3) SKILLS: IBM will work to help students, workers and citizens acquire the skills and knowledge to engage safely, securely and effectively with cognitive systems.
Operating under these principles, I have been fortunate enough to work with clients and partners in services or proof of concepts. These early engagements were critical in informing priorities and in understanding what researchers needed in order to be able to work independently with Watson. Any healthy relationship relies on trust. To help me trust Watson, it shows me evidence, what it has read, why it makes certain conclusions, and what specific entities, ontologies, relationships have been shared?
Watson for Drug Discovery embodies Design Thinking and how multidisciplinary teams can create a “beautiful product”1. Our design team has done an exquisite job of taking complex analyses, and simplifying and making them visually appealing, while maintaining traceability to various layers of supporting evidence.
Beauty and brains
An attractive product is one thing, but does it really work, and what is the evidence of efficacy?
Can Watson for Drug Discovery really generate new hypotheses, identify new targets, prioritize drugs for repurposing?
Yes – Pfizer generated new hypothesis for IO combinations
“Using this methodology, we have been able to identify new indications. We can assess gene targets we are pursuing for an immune response, as well as identify novel gene sets that have not been associated with an immune response that fight the tumor in a different way.”
– Dr. Paul Rejto, Head of Oncology Translational Research at Pfizer
Yes – Barrow identified and validated novel targets for ALS, never before associated with the disease
“By using Watson for Drug Discovery, we can make scientific breakthroughs in a fraction of the time and cost, increasing our knowledge of diseases faster than ever before.”
– Dr. Robert Bowser, PhD, Chairman of Neurobiology, Professor of Neurology and Neurobiology, Barrow Neurological Institute
Yes – University Health Network of Toronto identified drug candidates to treat Parkinson’s Disease
“The first of Watson’s six picks is undergoing lab testing. It’s a drug that’s already FDA approved, so it could quite easily be repurposed for Parkinson’s disease. And it’s showing some very exciting preliminary evidence in our animal models that it’s working.”
– Dr. Naomi Visanji, University Health Network of Toronto
Concept to consumption
Watson for Drug Discovery is trained in the language of researchers, automatically assembling a rich knowledge graph from literature, patents and drug/chemical database information2. We have created a scalable platform available to individual users, small biotechs and enterprise clients. Above all, we have made it consumable.
Tune in to the Discovery Channel at 9pm EST on June 21st to hear how Watson for Drug Discovery is helping repurpose existing drugs to treat Parkinsons.
If you are a biomedical researcher, I encourage you to put the power of 6000 research assistants3 at your finger tips. Apply here for your 30-day free trial, and start accelerating your discovery today!
(1) A statistically significant percent of users.
(2) Data from ClincalTrials.gov will be available soon.
(3) Research scientists can read about 300 papers a year, with about 1.8M research papers published each year. Watson can read them all, making it equivalent to approximately 6000 researchers at your fingertips.