An insight into artificial intelligence in drug discovery: an interview with Professor Gisbert Schneider

Introduction

As part of our special focus issue on artificial intelligence in drug discovery, we are delighted to interview Gisbert Schneider of ETH Zurich, Switzerland who shares his background and experience in the field of drug discovery and his outlook for the future.

We put the following questions to Professor Schneider:

Q1: What are your current research interests and how did you become interested in this field?

I’ve been studying adaptive biological and nature-inspired systems for more than 30 years, and still have only scratched the surface. I am convinced that nature provides us with solutions to most if not all our problems. After all, humans are a part of nature. All we have to do is look and observe without prejudice. Examples range from nature-inspired artificial intelligent systems, like neural networks models and evolutionary algorithms, to bionics and phytomedicine. My fascination with nature-inspired computing was greatly influenced by a book, namely ‘Parallel Distributed Processing’ by J. L. McClelland, D. E. Rumelhart, and the PDP Research Group[1]. I’ve been active researching at the interface between the life and the computer sciences ever since. Importantly, I also had excellent mentors during my university education who gave me the necessary freedom and leeway to explore unconventional ideas off the beaten path.

Q2: Your review discusses advances of AI use in drug discovery; do you feel that these advances have been duly recognized in drug discovery and development?

In fact, AI methods for drug discovery have been available for over two decades, but only recently have they become mainstream in medicinal chemistry. The reasons for that are manifold. For example, today we have access to faster computers and improved algorithms, much more chemically and biologically relevant data, and in particular widely accessible software libraries for AI modeling. Furthermore, research in the pharmaceutical industry has been, and to a certain extent still is, largely technology driven, with a prevailing mind-set for a certain period of time until the next technology wave is adopted. There actually are understandable reasons for this periodicity. However, until recently, AI applications were considered sub-par compared to experimental high-throughput screening, combinatorial chemistry, and other technological drivers. It was unimaginable that a computer algorithm could actually generate novel chemical entities with desired properties from scratch; possibly even better than a human expert. Meanwhile it has been realized that a machine intelligence can in fact positively support drug discovery and development. Certain algorithms partly cope with the cardinality of the chemical space and the nonlinear, often unknown multi-dimensional structure-activity relationships that govern the pharmacokinetics and -dynamics of a drug. However, we should be careful not to put all of our eggs in the machine learning basket, and look beyond the current hype around AI. There are amazingly few prospective drug discovery studies with AI. Most of the published examples are merely applications of deep learning methods lacking experimental validation. After all, at some point the computer-generated molecules must be synthesized, tested, and optimized.

Q3: You are a founding Director of the ETH (Eidgenössische Technische Hochschule) RETHINK think-and do tank – what is its mission/vision and what role will it play in advancing the use of artificial intelligence in drug discovery?

The RETHINK think-and-do tank was established at the ETH Zurich to bring together scientists from different disciplines and backgrounds to explore the opportunities and limitations of applied AI. We started with a first challenge a few years ago, on the topic of ‘rethinking drug design with AI.’ Meanwhile, the second challenge has been launched, namely ‘rethinking architecture with AI.’ The central idea of this initiative is to enable crosstalk between established scientific disciplines by offering workshops and other carefully guided events, and a space for engagement and open interaction. It is not very common to see philosophers, theoretical physicists, chemists, computer scientists, and architects from academia and industry discuss the drug discovery process. There are almost magical moments when these deliberations lead to new insights and inspiration. RETHINK is an experiment. We’re exploring new ways of interdisciplinary collaboration. I hope that this particular initiative will contribute to our understanding to which degree a machine intelligence can augment the creativity of medicinal chemists and support decision making in drug discovery.

Q4: What are the most prominent challenges the think tank faces?

In the beginning, the greatest challenge was to connect with interested individuals who were open to the special kind of scientific discourse the think-and-do-tank offers, and willing to invest the time. There’s no free lunch. Everyone who wishes to participate has the obligation to do this actively, take a step back to obtain a fresh perspective on one’s own research and scientific approach and, most importantly, constructively contribute to the events. Meanwhile, the RETHINK think-and-do-tank is thriving, and we have established a global network of committed partner organizations with which we share common goals. Patience, persistence and convincing events were the key. We aim to generate tangible outcome of each event whenever possible, for example by publishing a white paper summarizing the results of a workshop[2]. However, this initiative critically relies on direct interactions between people, which has been (and still is) largely curbed during the COVID-19 pandemic.

Q5: You have recently assumed the position of Director with the Singapore-ETH Center (SEC) in Singapore. What most excites you about this development and how do you see it affecting your future research?

The SEC offers an exciting research environment. It is a nonprofit research institution of the ETH Zurich located in Singapore, which was established together with Singapore’s National Research Foundation. The multifaceted activities of the SEC are focused on developing practical solutions for a liveable world. The SEC’s setup and unique diversity enables holistic transdisciplinary research through consequent scientific reasoning and creative thinking. To me, this setting is perfect, as it offers much inspiration for my own scientific research. The future is green and healthy. There is no alternative. This realization is a main driver of research at the SEC. The SEC project teams study the health of cities and their environments, networks and communities, and ultimately the health of the individual. Natural and artificial intelligent systems and design thinking form the methodological backbone to develop sustainable solutions for the world of tomorrow. We can receive much inspiration for our work from studying natural processes and solutions around us.

Q6: The coronavirus pandemic has provided numerous global challenges; how has COVID-19 affected your work?

I guess I speak for many by saying that I miss face-to-face encounters and stimulating conversation and discussion of scientific questions most of all. To me, there is no substitute for personal interactions to spark new ideas. Online video conferencing brings us only so far. On the other hand, the COVID-19 lockdown and especially working from home have helped me catch-up on some projects, and allowed me to dig deep without the constant distractions in the lab and office. Without the daily noise of the workplace the quiet of my home helped me focus and regain energy for the future. The pandemic has once again made it unmistakably clear that humans are part of nature. For my part, I’ve decided to devote much of my future research to my passion of nature-inspired drug discovery with and without AI.

Q7: Can you provide some examples of the role AI has played in the pandemic as well as how it could be utilized?

This is a question I have been asking myself a lot. I am surprised to see no therapeutic agents for COVID-19 at all resulting from application of this technology. Although theoretical papers (and startups) are mushrooming, touting the ‘power’ of AI for antiviral drug discovery, there are exceptionally few hit compounds reported. One can speculate why that is. Possibly, there still is too little data on COVID-19 targets and ligands available to train predictive data-hungry machine learning models. Possibly, we’re approaching the challenge in the wrong way. What we need are drug discovery methods and concepts for drug design in low-data scenarios. This is a worthwhile field of research.

Q8: What exciting developments do you foresee occurring in the use of artificial intelligence in drug discovery in the next 5 to 10 years?

My expectation is that improved prediction models for a variety of preclinical endpoints will be obtained from training machine learning models with combinations of theoretical molecular representations and easy-to-obtain experimental readouts, for example spectral or phenotypic data. This way, the current limitation of commonly used molecular representations can be overcome, and a link to biologically and pharmacologically relevant targets is obtained. We will also see the amalgamation of AI with lab automation and robotics for rapid feedback learning in automated design-make-test-analyze cycles[3]. The immediate availability of off-the-shelf software tools for deep learning will further the widespread application of AI models for drug discovery. Therein also lies a danger, namely the lure of hasty promises. Domain knowledge and expertise will remain quintessential for future drug discovery with AI. Because we’re interfering with adaptive biological systems, we can expect only partial predictability from the current mathematical models, especially when it comes to advanced preclinical and clinical endpoints[4]. My recommendation is to always combine deep learning with deep thinking.

Disclaimer

The opinions expressed in this interview are those of the interviewee and do not necessarily reflect the views of Taylor & Francis.

Declaration of interest

G Schneider declares a potential financial conflict of interest as a co-founder of inSili.com LLC, Zurich, and in his role as a consultant to the pharmaceutical industry. He has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This manuscript was not funded.