Cognitive Tech is the R&D Game-Changer
IBM Watson’s John Piccone on how their technology will transform pharma R&D.
2016 may be the year where Artificial Intelligence (AI) leaps firmly from science fiction into the present.
Recently, Google’s DeepMind AlphaGo program scored a landmark victory for AI, beating the world champion Lee Se-dol at Go, a game so fiendishly complex that the best players rest on attuned intuition to make decisions.
Since Alan Turing famously defined the tests for AI in terms of capacities that mimic human intelligence, it was inevitable that progress would be showcased by human-machine duels. But while landmark moments make for great public spectacle, like IBM Deep Blue beating grandmaster Kasparov at chess or the successor IBM Watson winning the popular game show Jeopardy, this performance obscured the real remarkable story.
In the build up to the event, Se-dol had been himself coaching AlphaGo to improve its play; although Se-dol is a prodigious Go player, he is not a computer scientist. Yet he was able to interact with the computer, and the computer was able to learn then generate new learnings.
This development is a key characteristic of a new form of computing that many are calling cognitive tech.
The traditional programs that most of us are used to were designed with completely different principles in mind, to execute a pre-defined recipe of actions incredibly quickly. While they were effective at this, they were very limited in their ability to discover new knowledge. But now, due to advances in both statistical techniques for working with big data and inference engines, the tools that store and apply logical rules to information, cognitive technology can make decisions and uncover completely new insights.
A reason this is so powerful is also due to improvements in Natural Language Processing (NLP), techniques that help computers understand the semantics and meaning in text. Progress in NLP has opened up whole new realms of data for analysis.
Most analysis is done with structured data, which conceptually are ordered into spreadsheet-like-databases. However, this is only 20% of the data out there. The other 80% of data, described as unstructured data, exists within all sorts of messy objects like word documents, PDFs, photos and videos.
Recent advances in NLP have basically unlocked much of this unstructured data. When combined with better computing and packaged so that non-technical researchers can interact with it, it has created an incredibly powerful tool for working with data and discovering new knowledge.
We spoke with John Piccone, the Global Leader for Life Sciences Offerings at IBM following his presentation at eyeforpharma Barcelona to understand how this transformation could radically impact pharma in just the next few years.
Q: Thank you for joining us John. It seems we are in a very exciting moment for healthcare. What are the biggest changes that cognitive tech could bring to pharma?
There are three big advances that IBM Watson technology can fundamentally enable.
We will see much more productive R&D, with more rapid development of novel therapeutics and in greater volume. Researchers will get the ability to ask questions and get answers more quickly. This will allow them rule out more blind alleys at an early stage and get straight to productive insights.
Secondly, once you’ve developed your early stage concepts, we believe that a more thorough and quicker research phase will create benefits throughout the R&D pipeline, and this basically comes down to fact that you are picking winners more accurately.
Finally, once you’ve developed the compound and it comes to approval, you will be able to use our cognitive functions to draw on much more evidence to show efficacy and health economics evidence.
“We used Watson Health’s discovery advisor and in just a few days found 20-30 new candidates. The confidence was so high in some selections that the client selected 8 of them for wet lab experimentation, which showed high activity in the desired properties. What would have taken decades took Watson Health just a few hours.”
Q: Can you talk me through a typical use case?
One engagement where we dramatically accelerated therapeutic innovation was with working p53 proteins, a known tumor suppressor. Modifying p53 and cross-correlating the results is an important technique for developing new cancer therapeutics, but it is incredibly labor intensive. Typically, just one high potential p53 kinase is encountered every few years.
We used Watson Health’s discovery advisor and in just a few days found 20-30 new candidates. The confidence was so high in some selections that the client selected 8 of them for wet lab experimentation, which showed high activity in the desired properties. What would have taken decades took Watson Health just a few hours.
Q: What were the mechanics of how this worked?
The Discovery Advisor tool is able to use our advances with cognitive computing to read texts from a vast body of published research and then extract the concepts. For example, it is able to read words and classify them to say these words are genes, these are proteins and these are disease names.
After the text is classified, it is able to identify the relationships and come to complex observations such as, this particular gene expresses this protein which tends to activate this biochemical pathway, which then relates to this disease.
Once these relationships are established the researcher is able to interact with this vast body of research by asking questions. If you wanted to find out what genes you should explore to produce a certain effect, or what compounds are likely to create a certain interaction, you could do so. It is this combination of opening up the technology to be used by an experienced researcher that makes it particularly powerful.
Q: Watson’s performance seems directly related to availability of information. What challenges does this create for you in an industry where there are many competitive and legal barriers that prevent people from sharing data?
We don’t see this as a critical obstacle, just something that needs to be factored into your project plan, as it will take time and effort to identify what knowledge is available in a particular domain.
We have groups within Watson Health that just work solely on this, and they have been really very effective. We obviously help companies get the most out of the proprietary data they have, but there is a huge amount that can be done with just knowledge in the public domain.
Q: There are many providers working in big data and also with cognitive tech, what is special about Watson Health’s approach?
We are building specific some specific applications that excel at a single problem, however, more importantly, we are building a platform of cognitive technologies that can be assembled by ecosystem participants – pharma companies, healthcare providers and payers to solve problems and provide solutions across the healthcare ecosystem.
Although there are many different tools for data analysis, processing, or visualization, we bring it into one place to create a seamless experience that enables transformative research workflows. By putting everything together and making it very easy for the researcher to ask questions and work with the technology, we think we can deliver huge benefits for R&D innovation, productivity and time to value.
One can imagine a world in R&D Pharma where every researcher has access to a Watson advisor which is fully informed with the world's corpus of scientific knowledge rather than just their own personal knowledge and can engage in dialog and answer questions to accelerate researchers' inquiries.
Q: Once systems are more embedded within pharma, and the capabilities develop even more, what does the future look like for pharma with cognitive tech?
Cognitive solutions have the ability to reason, learn and engage - with these capabilities, Watson health is targeting solutions that augment and scale human capacity and expertise. One can imagine a world in R&D Pharma where every researcher has access to a Watson advisor which is fully informed with the world's corpus of scientific knowledge rather than just their own personal knowledge and can engage in dialog and answer questions to accelerate researchers' inquiries. This will accelerate innovation in target identification, candidate selection, drug repurposing, clinical trial design and execution, safety surveillance and countless other R&D functions.
Since you're here...
... and value our content, you should sign-up to our newsletter. Sign up here