When it comes to molecular modeling, experimental scientists typically face two major barriers: misperception and trust. The most common misperception is that molecular modeling requires significant coding expertise or a Ph.D. in computational chemistry or theoretical physics. With regard to trust, it can be hard for scientists to believe in technologies that they have not used before, or they may have preconceived notions about the validity of modeling data. When modeling, it can appear that it would be easy to do something “wrong,” as there are many parameters and settings to take into consideration. These are reasonable fears for any good scientist to have.
With respect to the first barrier, in reality, there is a difference between the theoretical and applied aspects of modeling. Scientists involved in the theoretical aspects of modeling assuredly have a robust theoretical physics and chemistry background and are expert coders. Applied modeling, on the other hand, incorporates user-friendly tools and interfaces so that non-experts can get the benefits of molecular modeling without necessarily requiring a background in theoretical physics or expertise in coding.
Regarding learning to trust modeling, an analogy to consider is that most synthetic chemists use various tools in the laboratory, i.e. a nuclear magnetic resonance (NMR) or infrared (IR) spectrometer, without necessarily having expertise in the underlying mechanisms of the equipment. Rather, these chemists build skills in using and interpreting the results from the equipment, leaning on support as needed for complicated cases or continued education. The same can be true with molecular modeling.
My Own Journey
I started my career as an experimental synthetic chemist and later made the transition to computational modeling. I went to Columbia University for my Ph.D., where I studied the synthesis of main group metal catalysts for green chemistry reactions, such as CO2 capture and conversion.
The chemists that take the time to learn the skills now are going to be the most equipped for solving the next generation of materials science challenges
Many colleagues in this field of research use a trial-and-error approach to their research, relying largely on chemical intuition. Personally, I became interested in finding the most efficient, systematic approaches to discovering the best catalysts. This thinking is how I became drawn to computational chemistry. I felt that if computational chemistry could help me predict what might happen before I started to physically mix chemicals in the lab, then maybe I could be more efficient in my work. Columbia did provide access to computational tools, but at that time, I was a beginner, using modeling in only a few rudimentary ways. Also, at that time, running some simple calculations on metal catalysts was quite time and resource-intensive.
During my postdoctoral years at the Weizmann Institute of Science, I focused a lot of effort on learning how to incorporate molecular modeling into my research. I learned from doing, scouring the literature, making mistakes and talking to colleagues with more expertise. At that time, I was learning largely from the command line (preparing inputs and parsing outputs using a text-based terminal rather than a graphical interface). I was not aware that there are tools with extensive Graphical User Interfaces (GUIs). But, nonetheless, I was successful in my quest to understand how to model.
For me, modeling was a perfect complement to working in a wet lab, which I began to find more and more dangerous and tedious. I also came to appreciate that the value of modeling can come “before the fact.” Atomic-scale simulation can accelerate the development of new materials by helping to identify the most promising structures and formulations before you begin synthesis and testing. Despite this realization, I believe that in materials science, unfortunately, modeling is still very much used as an “after-the-fact” tool. In this way, the materials science field is behind the life science field. That’s actually what brought me to Schrödinger.
Value of Computational Modeling Tools
Today, one doesn’t necessarily need a Ph.D. in theoretical physics to successfully incorporate modeling into their research. At the same time, I always caution new modelers to not think of computation as a ‘black box.’ You cannot expect to get quality answers through modeling by pulling levers and pushing buttons without any understanding of the underlying physics and chemistry. Just like with using a spectrometer, there is a baseline amount of theory that is necessary and sufficient. That said, there are many tools now available to support modeling for non-experts. I’ll just highlight a few of Schrödinger’s that are somewhat unique:
- Comprehensive Graphical User Interface. The Materials Science Maestro GUI allows you to build structures and systems ranging from small molecules to complex formulations. Panels can then be employed to run simulations and analyze outputs, obviating the need for command line (of course, command line capabilities exist as well!) We have close to 200 panels in the GUI for materials science applications, significantly lowering the barrier for performing countless workflows.
- Educational Offerings. We maintain a comprehensive suite of educational materials that we are consistently enhancing. To date, we have close to 40 step-by-step tutorials that explain how to run different workflows using our software. These tutorials are written by our education team in collaboration with the expert scientists at Schrödinger most familiar with the application at hand. We also have online courses, which are fantastic ways to begin learning, especially because there are no hardware or software requirements involved, just access to high-speed internet.
- Support Services. Modeling is not something you learn in one sitting – you need to have a willingness to learn and take the time to do so. Schrödinger’s expert computational chemists serve as a guide during this learning process, providing in-depth support to non-experts as they get up to speed with molecular modeling.
With the many tools and opportunities available for learning, there is no reason that molecular modeling has to be out of reach for any scientist or company. It’s purely a matter of changing perception and using the right resources. I believe that it is only a matter of time until modeling becomes as ingrained in materials R&D as it is already in life science; the chemists that take the time to learn the skills now are going to be the most equipped for solving the next generation of materials science challenges.