Dr Zach Pitluk & Dr Srikant Sarangi on preparing the way for AI/ML to impact drug discovery and development
In the pharmaceutical industry, many see artificial intelligence (AI) and machine learning (ML) as the future of drug discovery and development. Big pharmaceutical companies are paying large sums of money upfront to AI specialists to kick-start programmes and, in addition, many have instigated in-house projects and technical developments, too.
But, even within this group, there is a consensus that the impact of these approaches is currently focused on chemistry (not biology), on small molecule candidate identification and/or triage and that it is still in its infancy with regard to biologics and biopharma development. Many believe that AI/ML is likely to deliver real value only in the long-term when biological, clinical and patient data are incorporated together. It is only when this is achieved that a more end-to-end idea of having AI suggest the next set of experiments can be realised.
However, companies understand that it is crucial to get started, and use what they can do currently to establish principles and practices. Some are already more bullish than others.
There is no doubt that AI/ML will play an essential role in all drug discovery and development going forward. Barbara Lueckel, head of Research Technologies at Roche, was recently quoted as saying: “I think it is fair to say we see machine learning as something that can no longer be disconnected from how we think about the continuum, from target discovery to clinical development.” Lueckel highlights that ML and AI are clearly high on the agenda for big pharma and something that cannot be ignored. But are pharmaceutical companies ready to fully embrace AI/ML in their labs? And how can they increase their return-on-investment to ensure their AI/ML-based platforms are making the biggest impact on their drug development efforts?
Data Defines Development – The Challenges
AI in healthcare is less about replacing the duties of providers and more about analysing the troves of data in clinical records to support better decisions. One challenge of using AI and ML is how to get lab instrument data into an analytical platform that can reduce time to innovation with ad hoc exploration and hypothesis generation. Typically, this data has been extracted but hasn’t been utilised if, for example, an experiment fails. And with drug development having an overall failure rate of 96%, there is a lot of data that could be inserted into an analytical platform for further interrogation to support a range of different trials.
The driver for change has come with the recognition that data has value beyond the initial experiment. It can be thought of as having money in the bank, which is where the urgent need is coming from and where data analysis can serve the most value – rather than being thrown into a data lake to be found at a later stage. This is the entire purpose behind the acronym FAIR (data that meet the principles of findability, accessibility, interoperability and reusability): to make data an asset that can be utilised.
Drivers Of Change
Another significant challenge for the lab of the ‘near’ future is how to take advantage of the expanded types and scale of data available – new analytical measurements, clinical trial data, and patient lifestyle information, for example. The more information an analytical platform holds, the more AI/ML has a potential role to play. Typically, computational tools have been too rigid to scale beyond a few patients’ worth of data, and this means researchers must extrapolate to population scale from just a few data points. Ideally, you need thousands to tens of thousands of individual patients before you can confidently identify important, population-relevant features within a dataset. This is where AI and ML can be drivers of change, improving the analysis of large and complex datasets, maximising staff efficiency, and reducing full-time equivalent (FTE) time through less data wrangling and more automated analytics.
Solving these challenges with a high-performance integrative analytics platform that enables storage of complex multimodal data, easy-to-use ad hoc computation and analysis capabilities, with a predictable cost model, can support the adoption of AI/ML and prepare companies looking to advance their drug discovery with these new approaches.
It is data that transforms hypotheses into knowledge and knowledge into innovative discoveries. With the massive development efforts on AI/ML, the way discovery and developments are made will certainly change, with the pharma industry expecting it to deliver better drugs more quickly and at lower financial risk.
Ahead of this ‘end game’, however, high- performance computing platforms and storage solutions are vital preparation.
Dr Zach Pitluk & Dr Srikant Sarangi are with Paradigm4
