Computerize Drug Discovery - Start at the Beginning
- How the pharma industry can utilize advances in digital technology to solve old problems -
Wolfgang Kissel and Herbert Treutlein, 15 May 2018
The pharma/biotech industry is under mounting pressure to increase its R&D productivity. Some, mainly bigger, companies cut early R&D and leave this risk to start-ups and smaller companies. They then buy them or their projects when the risk seems to have decreased. Others state they want to utilize advances in digital technology, become more agile and decide to team up with Silicon Valley and focus R&D on data and digital.
To computerize key processes in the pharma industry is still a major enterprise, which only the brave dare to start. Now, Novartis’ new CEO Vasant Narasimhan has stated that data and digital are a part of his new strategic plan for Novartis. By taking references to the software industry’s agile project management, he signals internally that he is serious about it.
However, looking at the broader reality we see that, despite best efforts, the industry is struggling to make way for real innovation. Although on the forefront of medical development, the pharma industry is deeply conservative in its structure and culture. In a recent article in the biotech newsletter Endpoints, the author describes the struggle of R&D to change when searching for an Alzheimer’s drug (1). Although late stage failures in Alzheimer research are happening one by one, the industry continually repeats studies on the amyloid beta target. This is just one example. It seems hard to give up old habits and changes are only slowly adopted.
Some, like Genentech, have built up computational drug discovery groups and seem quite advanced in their digital technology applications (2,3). Still, new technologies and approaches seem mere add-ons to support old ways. In our own field of computational drug discovery and design we observe that none of the bigger pharma companies apply a consistent approach to discover and design drug candidates from scratch with digital means. Others, including us, have shown that it is possible and that R&D productivity can improve dramatically when computerizing the key processes in early drug discovery (4,5).
From our work, we see two main challenges to computerize R&D, especially in early drug discovery. First, it is to innovate the key processes that will create the biggest leap in productivity. Second and even more challenging is to fundamentally change the mindset of those involved in these processes. Acquiring new technologies alone or teaming up with Silicon Valley will not achieve the productivity gains that are hoped for.
Start at the beginning
Computerizing early drug discovery has to start at the very beginning of the process and not as support for the existing approach. We have worked successfully with this premise for many years now. It is our experience that the necessary fundamental process redesign is mostly neglected. Concurrently, the people involved are not expected to question their role and self-concept. This leads to shallow changes labelled with buzzwords rather than harnessing the superior productivity potential of computerizing drug development.
From our experience, to computerize early drug discovery processes allows you to “Do it right the first time”. It avoids dragging along and culminating the limitations and mistakes of “trial and error” into the clinic, often leading to late stage failure. Experimental methods should only be applied for testing and verification, not for molecule discovery and design. Computational drug discovery can do this much more accurately, quicker and cheaper, so that drug candidates can be built, like on an assembly line.
In our approach, we have turned the current drug discovery back on its feet by being target and software driven. We don’t accept the limitations of wet chemistry. We start with the computational exploration of the target and then virtually screen against molecule libraries. Or, with our multi-fragment molecular modelling approach, we find the best fragments that could be the basis for designing a promising scaffold. We use biological assays as support and help for verification, and wet chemistry for the synthesis of the predicted molecules. With our iterative process of design, testing and evaluation, we apply a similar approach as software companies conduct with their concept of agile.
We haven’t yet seen a bigger company that has put their computational drug discovery in the strategic driver seat and has developed a new drug candidate computationally from scratch. We believe that only the consistent application of computational drug discovery beyond "help and support" will fully exploit the potential of computerized early drug discovery where the value drivers are:
▪Accuracy – where targeted design replaces the still widely applied guesswork. The process of trial and error changes from testing hundreds/thousands of compounds to testing only small focused molecule libraries with the best predicted properties
▪Speed – where the process gets accelerated >5x quicker than the conventional process, shortening the process of target validation to active compound from years to months/weeks by replacing most of the costly and time consuming wet chemistry
▪Cost – where the process of target validation to active compound becomes >4x cheaper and reduces cost from $$$M to $$$K
▪Risk mitigation in early drug discovery from costly failure in later stages to eliminating risk at earliest possible stages.
With these value drivers we achieve >2000% process improvement. We expect that this will result in patentable compounds in 12 months.
The even bigger challenge to achieve the above-anticipated quantum leaps lies in changing the mindset of all those involved. To do it right, computerization needs to be the fundamental driver and not just a tool for incremental improvement. This requires, besides the above-mentioned re-invention of early drug discovery, a new self-concept of the parties involved.
The vast majority of pharma/biotech companies are stuck in an old paradigm where “traditional” medicinal chemistry is the ultimate driving force in drug discovery. New technologies are seen as add-ons to support and improve the old paradigm, but not to change it. Again, our experience shows that traditional medicinal chemists often act as the impermeable layer, that prevents the large-scale adoption of new technologies like computerization. The necessary key changes we see are:
▪Defining the chemist as a supporter rather than the driver, whose capabilities otherwise become the limiting factors in the process.
▪Accepting that molecules are a special kind of information, which can be found and made, and that chemistry is a useful instrument.
▪Subordinating to a strict and structured process management with minimum waste.
To learn this new concept in the least threatening way, we believe that an experimenting approach with real projects works best. An organization can learn and adapt without disrupting its current processes. With the experience of the new concept, the organization can then make the switch and leave its old ways behind.
All this puts the scientist and her creativity back to the centre of the process. And with the new value drivers this creates a very real probability for the productivity revolution that the industry strives for.
P.S. In case readers are present at the Bio 2018 Convention in Boston in June, we could meet and discuss and exchange our experiences.
How to start
You might wonder that many articles like this describe well what one must do to achieve a certain goal. Although the what to do is important, the how to do it is much more essential. We therefore like to add some strategy for how to start such changes.
Introducing digital technologies, wherever, is a major change that affects all areas of the business. Introducing digital technologies in R&D might be the trickiest area, given that R&D considers itself as the forefront of innovation in pharma/biotech. In any case, we are talking about a major change Vasant Narasimhan and other CEOs want to introduce to their organizations.
With the recent announcements about to computerize R&D, there is a high risk that the new paradigm of data and digital gets imposed on the organization, foreshadowing failure or at least significant slowdown.
From our experience in large-scale change, change of such magnitude is usually considered as a threat. When the new strategy gets announced, nobody will speak out against it because who is against change? But when the people affected realize that their way of working, and even their role, is fundamentally questioned, they naturally want to prove that, besides some cosmetic changes in the manner of support and help, the fundamental change is not necessary.
The best way to start is not to immediately change anything large-scale, but to begin by establishing small organizational units where all planned changes are implemented at once without compromise. This may sound counterintuitive knowing that CEOs want to make such changes across the board and do so quickly. However, setting up such organization units allows you to experiment with and test the new paradigm on real projects. This is the least threatening approach.
We suggest starting with early drug discovery, because it is the start of the whole drug development process. It allows you to learn the new concept and adapt. It allows the people involved to get used to the change in their role and to experience the quantum leaps in productivity through computerizing the early steps in drug development.
We also suggest selecting the most curious and “adventurous” people in R&D. This organizational change follows the same principle of any innovation. The early adopters and fast followers establish the innovation and the rest will follow when the benefit has become obvious. To be completely honest, you should also drive this change to lose the laggards along the way. This is only healthy because they are the ones who will slow down or resist even the softest change.
Still, some people with power and resources will be tempted to impose this change on the organization; but we can almost guarantee this will be a lengthy and costly exercise. The approach of change we suggest might take longer at the beginning but is sustainable. With the new value drivers, it creates a very real probability for the productivity revolution the industry strives for.
(2) Tsui V, Ortwine DF, Blaney JM. Enabling drug discovery project decisions with integrated computational chemistry and informatics. J Comput Aided Mol Des. 2016;31(3):287-291. doi:10.1007/s10822-016-9988-y.
(3) Ortwine DF. Computational Support of Medicinal Chemistry in Industrial Settings. Methods Mol Biol. 2018;1705:345-350. doi:10.1007/978-1-4939-7465-8_16.
(4) Abel R, Mondal S, Masse C, et al. Accelerating drug discovery through tight integration of expert molecular design and predictive scoring. Curr Opin Struct Biol. 2017;43:38-44. doi:10.1016/j.sbi.2016.10.007.
(5) Currier MA, Stehn JR, Swain A, et al. Identification of Cancer-Targeted Tropomyosin Inhibitors and Their Synergy With Microtubule Drugs. Molecular Cancer Therapeutics. May 2017. doi:10.1158/1535-7163.MCT-16-0873.