Science & Innovation

To Infinity and Beyond – The Parallels between “Toy Story” and Drug Discovery

Animated movies have served as a dominant form of my family’s entertainment for more than a decade, so we were thrilled with the recent release of “Toy Story 4”. Our list of favorites is long — but the original Toy Story holds a seminal place both in our family’s memory and in film history; released in November of 1995, it was the first full-length computer-generated imagery (CGI) animated movie.

I am inspired by the innovation on display in the latest installment of the “Toy Story” quadrilogy for professional reasons too: There appear to be striking parallels between the maturation of CGI over the last quarter century and an ongoing revolution in the world of computational drug discovery. That may seem like a far-fetched statement. What could the animated antics of a floppy toy cowboy have to do with biomedical innovation? The answer: Physics. More specifically, the application of computer-powered, physics-based approaches to visualize and model the complex interactions of particles, objects, molecules, and structures in three dimensions.

To play out this analogy, let’s go back in time to the first attempts to capture moving images, the early predecessors of animation and “Toy Story.” The praxinoscope, which used a spinning cylinder to create a brief animation, was introduced in 1877. The first flipbook, also called a kineograph (Latin for “moving picture”) was patented in 1868.

These early technologies were very quickly replaced by ever-more sophisticated devices, including a moving picture camera, or “kinetoscope,” pioneered in Thomas Edison’s lab. The first audience to see a public showing of a motion picture was in France in 1895 — exactly 100 years before the box-office-shattering debut of the original “Toy Story.”

Before “Toy Story,” CGI had been used for short scenes (the “Genesis Effect” sequence in “Star Trek II: The Wrath of Khan”) at Lucasfilm, but it took Steve Jobs’s vision and determination — not to mention, his considerable investment — for the Pixar Image Computer and CGI to come into their own as a truly disruptive technologies poised to revolutionize the world of film. Jobs, who clearly saw the transformative potential of CGI, supported Pixar Animation Studios, which brought us that first Toy Story movie and so many other gems.

Over the next 25 years, the ability to render characters and scenes in 3D grew exponentially. As computing power increased, masters of the art of CGI found they could apply the principles of classical mechanics (i.e., physics!) to examine and then build gorgeously realistic models of people and objects in motion and layer simulations of their interactions with the environment. You can’t watch recent releases like “Moana” without marveling at how realistically CGI is able to render flowing hair and rippling water.

Returning now to drug discovery: The vast majority of small molecule therapeutics achieve their desired effect through reversible binding to a relevant protein target, but protein-ligand binding — which mediates all biological processes — is highly complex. In general, the pharmaceutical industry has approached therapeutic compound design empirically — meaning that for decades drug discovery scientists like me had to rely on solving multi-parameter optimization problems using expensive and time-consuming trial and error in the lab.

Synthesizing even a single novel compound may take weeks of skilled medicinal chemists time and cost thousands of dollars. This in turn leads to even well-funded drug discovery projects typically synthesizing fewer than 1,000 novel compounds a year. A predictive approach — by which we could determine ahead of time the best molecules to synthesize, and then focus the talents of chemists on synthesizing only the most valuable compounds that possess the desired features — would be an important step forward. This has long been the promise of structure-based drug discovery, but much like computer animation, progress was measured in the smallest of steps, and it has only in recent years lived up to its full promise.

Structure-based drug design — that is, using structural knowledge of the target and compound to estimate and optimize binding — was severely limited by the fact that we had only static images with which to examine the binding of ligands to protein binding sites. As an industry, we have had some terrific successes to date using empirical and structure-guided approaches, as evidenced by the number of life-saving medicines available today. However, many promising targets have been deemed “undruggable” or “intractable” because we did not have the capability to model the intricate “molecular dance” that occurs when compounds bind to proteins. We were in essence designing drugs based on static drawings of the binding site — we hadn’t even arrived at the rudimentary “flip book” phase of moving image evolution.

Schrödinger set out to change that in the early 1990s by developing software that could model the dynamic interactions between molecules and their cognate proteins. Our early investors, Bill Gates and David E. Shaw, believed in the transformative potential of computer-based modeling in drug discovery. Like Steve Jobs with CGI, they saw past the clunky first iterations to a future where realistic, physics-based modeling could transform an industry.

It was a good thing they were patient. Accurately modeling molecular mechanics and dynamics turned out to be an astoundingly difficult problem to solve. Using the work of Richard Freisner, a Columbia University professor and co-founder of Schrödinger, as a starting point, large numbers of scientists including physicists, computer scientists, and crystallographers struggled for more than a decade to develop simulation tools. We can now visualize proteins — essentially complex molecular machines — as well as how compounds move within a specific binding pocket of a protein and interact with water molecules (see video above for examples).

It turns out that modeling the behavior of water is relevant in both CGI animation and drug discovery. In drug discovery, atomistic modeling allows teams to calculate the energetic contribution and binding effects of water molecules. Designing optimal drug compounds is partly based on their ability to displace or retain specific high-energy water molecules (shown in red the images) in protein binding sites.

During the generation of animation effects, computer scientists and artists create realistic water by breaking it down into millions of tiny particles and simulating their movement and acceleration using a computer program. In both cases, a scientist’s ability to accurately simulate minute events at enormous scale enables them to create highly impactful designs that would not be possible without programmable physics-based models.

A key breakthrough came in 2012, when Schrödinger finally combined alchemical protein-ligand binding free energy calculations, an accurate force field (sounds like something from Star Wars, right?) for drug-like molecules, sampling algorithms, and one of the largest GPU-based high performance computing clusters (up to 100x faster than traditional CPU-based approaches). The confluence of these theoretical, technical, and practical advances enabled in silico simulations comparable to experimental assay results to contribute to real-life drug discovery efforts. Calculating binding free energy and understanding the molecular dynamics at play is crucial to predicting the potency and relative selectivity of a therapeutic compound.

Chemical space, defined as all possible small molecules including those involved in naturally occurring biological systems, is estimated to be 10⁶⁰ — which should be read as a one followed by sixty zeros — and is a number so large it might as well be infinite. To date, only a tiny fraction of this space has been explored in the context of medicine development. Physics-based predictions coupled with protein structure-function insights can accelerate the exploration of chemical space as well as facilitate the discovery of new bioactive molecules to explore our understanding of biological processes and new medicines to treat disease. Running these “computational assays” coupled with machine learning protocols utilizing this computationally generated data facilitates much wider exploration of chemical space than is possible using traditional methods.

The most encouraging aspect of the parallel between CGI and drug discovery is that advances made by one group have, in both cases, spurred tremendous advances across the wider industry. RenderMan, the software used to produce Pixar feature films, is now available to any animator, which has led to an explosion in the number of sophisticated and visually stunning films from multiple studios. In the same way, biopharma companies and academics around the globe are increasingly deploying Schrödinger’s drug design software platform to explore new corners of chemical space and accelerate in silico evaluation of compounds so that they can be confident they’re moving only the most promising into the more expensive phases of synthesis and wet lab analysis.

The reach of these two computational platforms has also been extended by publications and collaborations. For example, the original Toy Story and subsequent animated hits emerged from the now multi-decade collaboration between Pixar and Disney teams. Pixar in a box delivered by Khan Academy explores the science, technology and people behind animation.

The discovery of GS-0796, an Acetyl CoA Carboxylase (ACC) allosteric inhibitor for nonalcoholic steatohepatitis (NASH), now in Phase II at Gilead, is metaphorically the “Toy Story” equivalent of Schrödinger’s computational drug discovery platform. The compound originated at Nimbus Therapeutics, which was founded in 2009 by Schrödinger and Atlas Ventures. GS-0796 was the first clinical compound designed by Nimbus and Schrödinger using physics-based chemical property predictions. It was identified by scoring molecules for their ability to maximize the displacement of high-energy water molecules while simultaneously maintaining drug-like properties. Many additional drug candidates based on the technology have been discovered since then through numerous collaborations. Morphic Therapeutic and other biotech companies soon to emerge from stealth mode, have built discovery programs around our physics-based molecular modeling technology.

As with CGI-enabled 3D animation, advances in predictive computational chemistry software took patient investment, computing advances, and a lot of persistence, interspersed with flashes of brilliant insights. And just as CGI has revolutionized film, physics-based molecular modeling coupled with increasing access to structural data (e.g the emerging growth in Cryo-EM) is now truly beginning to revolutionize drug discovery.

While the juxtaposition of medical breakthroughs and animated films may seem odd, I see interesting parallels. Both the arts and medicine have historically had a huge impact on society. Coupled with technology, they continue to have enormous potential to improve our lives and make our time with family more enjoyable.

“To infinity and beyond,” the catch phrase made famous by Buzz Lightyear in “Toy Story,” seems to serve as an apt call to action: inviting drug discovery researchers to continue working on what were once unattainable solutions until they are at last within our reach. By combining the emerging application of vastly increased computer power, physics-based algorithms and machine learning with human creativity, ingenuity and insight, we are poised to open new worlds of possibility for drug discovery and to further our collective impact on human health.


Author Photo: Karen Akinsanya, Ph.D.

Karen Akinsanya, Ph.D.

Karen Akinsanya, Executive Vice President, Chief Biomedical Scientist, Head of Discovery R&D, joined Schrödinger in 2018. Karen leads our Discovery R&D group with responsibility for preclinical drug discovery and translational research. She has more than 25 years of experience in academia, pharmaceutical R&D, partnerships, and licensing. Karen joined Merck Research Labs in 2005 and held positions of increasing responsibility in clinical pharmacology as a development team leader working on first-in-human studies through late-stage label studies before joining Discovery Preclinical & Early Development as a therapeutic area lead and then a search and evaluation lead in business development. Karen received her Ph.D. from the Royal Postgraduate Medical School at Imperial College in London, in endocrine physiology. After post-doctoral training at Imperial and the Ludwig Institute for Cancer Research (UCL), Karen joined Ferring Pharmaceuticals in R&D working across sites in the UK and US. At Ferring, she led the discovery of a family of dipeptidyl peptidases related to DPPIV and pre-clinical characterization of FDA-approved FIRMAGON® for prostate cancer.

Sign Up Today

Be the first to know what is coming next.