Error, Nonlinearity, Incompletness
This year’s Digital Festival Zurich provided the stage for a special session “Heath’s Digital Future”, organized by ETH Zurich, and chaired by Shalini Trefzer. RETHINK director Gisbert Schneider discussed the opportunities and limitations of drug design with artificial intelligence (AI). According to Schneider, the world of chemistry is largely based on explicit knowledge as encoded in written text, molecular models, and in the formulation of underlying physical principles. Medicinal chemistry, and in particular drug design, faces a particularly challenging situation which, to date, has precluded the possibility of exhaustive problem representation using chemical terminology. The systemic pharmacological effect of a drug is governed by poorly defined, and highly-nonlinear, relationships between many contributing factors, leading to often unpredictable system behavior. In other words, medicinal chemists work ‘at the edge of chaos’ – with or without AI. Read more about this.
Sciences and medicine will converge
RETHINK, July 10, 2019. Dr. James Bradner, President of the Novartis Institutes for BioMedical Research (NIBR), visited ETH Zurich. He gave an inspiring ETH Global lecture, co-hosted by RETHINK. This highly recognized lecture series provides an opportunity to discuss the pressing questions of our time and learn from distinguished leaders in their field. Addressing the audience in the fully booked Siemes Auditorium, Dr. Bradner explained what it means for NIBR to shape the future of healthcare and charting a new course for collaboration with academic partners around the world.
The Mix of Mind and Machine
We are witnessing a renaissance of AI in all areas of medicinal chemistry. The previous wave of excitement crested roughly 30 years ago with the first applications of neural networks, inference systems and other machine learning models. After much debate, and too little in the way of convincing applications, the scare was quickly over again. Now the spectre of AI is back, this time, apparently, for good. The combination of laboratory automation and innovative software solutions for process planning and drug design promises better drugs, discovered and delivered faster. Medicinal chemists are exploring, adopting and adapting this technology. The prospects for this mixture of machine and mind in drug discovery are huge.
See the latest comment by Gisbert Schneider in Nature Machine Intelligence 1, 128-130 (2019).
Image credit: Jack Burgess
Bridging Between Scientific Disciplines
ETH Zurich, March 28, 2019. The President of ETH Zurich, Professor Dr. Joël Mesot, officially launched the RETHINK initiative in the fully occupied Siemens auditorium on the ETH Science City Campus. Outstanding speakers from leading international research organizations, including the eight RETHINK partner institutions, shared their thoughts on “design with artificial intelligence”. Their reports on cutting-edge discoveries at the interface between the life sciences, engineering, bioinformatics, architecture and philosophy fascinated everyone in the audience. RETHINK already succeeded on the first day of its mission to provide a top-notch forum for open dialogue between the scientific disciplines. Stay tuned for the next RETHINK event!
In a RETHINK seminar on March 29, 2019, Paul Wrede of Charité Universitätsmedizin Berlin defined intelligent behavior as the ability to solve problems and “to adapt to change”. Is intelligence only bound to a complex nervous system in higher organisms? This question may represent an unjustified anthropocentric view on the world, centric to the problem of intelligence in biology. Careful observations of small animals like the tiny fruit fly Drosophila melanogaster have revealed intelligent decision making. This may not come as a surprise because Drosophila possesses a central nervous system. Even single cell protozoa without a nervous system, such as ciliates, show clever and fast escape behavior.
What about plants? In fact, there is convincing evidence supporting the intelligence of plants. This includes physiological features that for a long time have been attributed to animal organisms only, such as the ability to see, taste, hear and smell, and use this sensory input to competently navigate in the environment. It begs the conclusion that an organism can be intelligent without neurons.
When discussing questions in biology, Theodosius Dobzhansky’s imperative “Nothing in biology makes sense except in the light of evolution“ stands above all. Evolution according to Charles Darwin describes adaptive processes within the living world. Adaption in biology means solving unforeseen problems and difficulties. According to Wrede, evolution can thus be construed as transient renewal. In other words, intelligence is evolution in very short periods of time. The difference between evolution and intelligence lies only in the time range.
RETHINKING Drug Design Mini-Symposium in San Francisco
Hosted by swissnex San Francisco, RETHINK fellows and a select international group of scientists presented the latest advances of AI in drug design. The audience engaged in a lively discussion about the applicability of these new technologies. Will AI deliver new drugs faster? Will AI help develop personalized medicines? What are the limitations of AI in healthcare? These and many more questions were actively addressed. As the first of its grand challenges, the RETHINK initiative at ETH Zurich strives to explore the possibilities and limitations of AI in drug discovery .
RETHINK fellow Dr. Jan Hiss presents the “Molegram Explorer” app for the HoloLens
Accompanied by a theatrical performance, RETHINK fellow Jan Hiss took the participants of the WORLDWEBFORUM on a journey to the world of molecules. During the performance, colorful holograms of molecular structures hovered on stage and mesmerized the audience. The Molegram Explorer is a “mixed reality” application for HoloLens glasses. This new technology helps both students and researchers alike to improve their understanding of 3D molecular surfaces and the basic concepts of ligand-receptor interaction. This project was realized with support of the LET department of ETH and the company afca. ETH Zurich is a scientific partner of the WORLDWEBFORUM. The annual meeting of outcome-oriented executives took place in Zurich on January 17, 2019.
Mind & Machine in Drug Design
Integrating domain-specific machine intelligence in the pharmaceutical industry is a moonshot program for AI in healthcare. The challenge is not only to generate novel drug candidates that are — practically — optimal in terms of their pharmacokinetic and -dynamic properties; the integration of AI will also require significant investment of time, money and the reorganization of laboratory structures and discovery processes. In consequence, the envisaged automated drug design engine may not only imitate but exceed human decision making as a core aspect of the drug discovery process. If successful in the long run, the approach will combine a continuously learning, chemistry-savvy AI with the synthesis and testing of pharmacologically relevant chemical matter. Academic institutions and not-for-profit organizations can offer the necessary leverage, and leeway, to explore unconventional thinking and challenge machine intelligence models to generate novel drug candidates. The RETHINK “Think-and-Do Tank” provides such an environment at ETH Zurich.
Read more about this exciting concept in the latest publication by RETHINK director Gisbert Schneider: Nature Machine Intelligence 1, 128-130 (2019).
Credit: Jack Burgess
RETHINK Workshop, San Francisco, December 5-6, 2018
A unique workshop on the topic rETHinking Drug Design with Artificial Intelligence took place in the Art Gallery “The Laundry”, located in San Francisco’s exciting Mission District. Twenty senior professionals and executives from industry and academia with different backgrounds – ranging from classical medicinal chemists, physicians, and machine learners to philosophers and architects – gathered for this first-in-class event. The participants aimed to identify those challenges of the pharmaceutical design process that might benefit most from applied artificial intelligence. Cross-disciplinary thinking in an open atmosphere was stimulated and enabled by the innovative format of this workshop. The proceedings will be prominently published as a white paper.