Paul Denny-Gouldson, VP Strategic Solutions at IDBS, explores what the future holds for drug discovery laboratories.
A team of researchers at the University of Manchester this month published promising findings on 'Eve' – an artificially-intelligent ‘robot scientist’ capable of screening potential drugs almost completely independently. Eve has discovered a compound shown to have anti-cancer properties, which might also be useful in the fight against malaria. Meanwhile, IBM’s artificially intelligent (AI) Watson system is seeing take-up among pharmaceutical companies. Is this how we can expect the lab of the future to materialise?
As the relentless march of technology continues and the laboratory becomes more automated, these types of systems will increasingly pool knowledge and data together to discover unique links – between otherwise seemingly unconnected observations. This couldn’t be timelier, with research and development (R&D) teams under intense pressure to develop new products faster and cheaper. New breakthroughs are becoming become scarcer and costlier – and we’re rapidly approaching the one billion US dollar mark for the development of a brand new drug.
The shift towards automation and robotics in the lab has been a long time coming. In the 1990s, many cited AI – or computational chemistry and molecular modelling, as it was then, and total laboratory automation as the ‘next big thing’ in R&D. But while some industries are seeing complete job displacement brought about by new technologies, the pharma sector doesn’t yet see that entire ‘replacement’ observation. Drug discovery labs, for example, still need a human element for the foreseeable future – and in my opinion that’s a long foreseeable future.
The role these AI systems play lies in helping explore fresh ideas, such as a new pathway or target for a drug, and they can offer this coupled with predictive analysis to supplement the work of scientists in the lab. The rise in companies offering predictive ADME/toxicology modelling is a prime example of this. These predictive analyses have limitations – the volume of data and computing power required to simulate the effects of a drug on the whole body are enormous, for example. But their empirical basis means they can now get quite close to ‘nature’ when the system is well defined and understood – i.e. when they have enough data to help build the model.
While that stage may still be some way off, our ability to manage and analyse more and more data has increased exponentially. When these new systems are performing any kind of analysis, they will have to feed back the derived information and integrate it with other new and potentially unrelated data. In that sense, a gradual adoption of more intelligent lab technologies will make it even more important to be able to securely manage multiple streams of data, from one central point of knowledge and IP.
It’s also worth considering the commercial practicalities of a ‘robot scientist’ system. Historically, the need to connect a network of computers made it an expensive option, viable only for the larger pharmaceuticals companies. However, with the advent of cloud elastic computer resources, this issue is no longer a barrier to adoption. It’s now about what makes sense to analyse and, more importantly, what questions do we want to ask? Putting all the data in to a ‘data lake’ and allowing these AI tools to crawl over it is one thing – getting them to do it in a directed and reasoned manner to answer specific questions is another.
Will the introduction of AI herald a new era for life sciences? Not quite – Eve is the latest evolution of this kind of technology, rather than a brand new concept. This is not a silver bullet but, as science evolves and the data availability changes around it, AI technology will play a role in the discovery and development of new drugs and disease understanding. As with all new tech and tools in sciences, it will help support the scientist – but it will not replace them.