(4av) Computational Design of Bimolecular Self-Assembly and Adsorption Behaviors through Thermodynamically Consistent Multiscale Modeling

Bio-based, and, in particular, protein-based materials are increasingly relevant as we transition towards renewable energy sources and sustainable feedstocks that reduce environmental impact. Novel biomolecules with unique self-assembly behaviors hold promise across many industries, from consumer products to medical devices and biopharmaceuticals. Crucial to understanding and engineering interactions between biological molecules, however, is the behavior of solvent, in particular water, which is not easily captured with coarse-grained models. Such difficulties arise from the intrinsically many-body nature of solvent-mediated interactions, such as hydrophobic associations. To address these issues, my research program will leverage fundamental synergies between statistical mechanics, molecular simulation, and information theory to enable the computational design of separation processes and soft materials.

A common motif throughout my research will be the development of design algorithms that couple optimization and simulation to explicitly account for and manipulate solvent-mediated interactions. Research activities will target 1) development of coarse graining and enhanced sampling techniques that accurately model the effect of solvent in bimolecular self-assembly, 2) improvement of free energy calculation techniques for solvation through coupling to recent developments in information theory and machine learning, and 3) application of these techniques to the design of membrane-based and chromatographic separation processes for novel biological molecules. The structure of the research program reflects the essential development of fundamental theory with specific applications in mind, as well as broad collaboration, which is key to the development of graduate students, especially those who pursue computational research. Theoretical and algorithmic developments will build a foundation on which my group will design soft materials computationally to meet precise societal and industrial needs. In particular, our methods will enable the rapid development, through computational screening, of both materials and protocols for the separation of proteins and other biomolecules. Agile development of these tasks will become critical as these molecules gain greater prevalence within consumer products and as materials and reagents, such as enzymatic catalysts, in industrial processes.

The proposed research will draw on my extensive molecular simulation experience, even dating back to undergraduate research, in exploring the behavior of water at interfaces and in designing novel simulation algorithms based on state-of-the-art machine learning techniques. During my PhD, I demonstrated how the response of water to solutes, in particular the modification of three-body angle distributions, could be used as a probe for solute free energies of solvation and association with interfaces (ACS Nano, 2017 and JCP, 2019). Through a genetic algorithm, I also designed the chemical patterning of interfaces to modulate dynamics of water, as well as binding affinities of a range of small solutes (PNAS 2018 and 2021). Analysis of optimized systems revealed fundamental connections between water entropy, dynamics, and structure, identifying ways in which surface interactions could be tuned to selectively influence these related quantities.

My postdoctoral work has explored modern Bayesian methods for coarse-graining and enhanced sampling (submitted to JCP). If we seek to not only engineer mechanical properties of self-assembling materials from a fundamental level, but also sensitivities to thermodynamic conditions, it is imperative that multi-scale modeling techniques be developed to ensure thermodynamic consistency. I have shown how training a variational autoencoder (VAE) simultaneously solves problems of coarse graining, enhanced sampling, and back-mapping, with various molecular simulation techniques appearing as special cases. Remarkably, a trained VAE may be used to learn MC moves passing through the coarse-grained space that satisfy detailed balance and exactly preserve the fine-grained ensemble, enabling the recovery of thermodynamic information across model resolutions. These results excitingly lay the groundwork for seamless switching between implicit and explicit solvent. Simulations will be accelerated through VAE-based MC moves utilizing coarse representations, while water crucial for the accurate representation of solvent-mediated effects can be introduced to maintain accuracy of thermodynamics associated with self-assembly or binding at interfaces.

Teaching Interests

As a general teaching philosophy, I strive to imbue students with an appreciation for both the history and practical applicability of a topic. I want students to internalize material such that they feel amazement as they work through classic problems and simultaneously question the subtle assumptions along the way. Students should gain both a healthy respect of the past, but also a skepticism and awareness of limitations. This view has arisen out of my experience as a graduate teaching assistant, including development of course material and guest-teaching numerous lectures, as well as designing and delivering a series of five two-hour lectures for high-school students through the School for Scientific Thought (SST) program, part of the Center for Science and Engineering Partnerships at UCSB. My course, “Modeling the physical world: gases, liquids, proteins, and everyday life,” developed students’ skills in creating and evaluating mathematical models of physical phenomena, which is essential to a STEM career. Classes consisted of both guided group activities, as well as interactive lectures. A component of participating in the SST program included participation in multiple seminars and lectures led by experts from the Girvetz School of Education at UCSB, as well as creating (and discussing) metrics to assess student engagement and understanding for each lecture.

My experiences as a teaching assistant and in the SST program have led me to an example-heavy teaching style and commitment to generating class discussion. The former is necessary to provide students an adequate mental picture of esoteric topics in thermodynamics or mathematics and helps to diversify the content-matter of the course. In-class discussion is critical for creating engagement amongst students, and also forces students to evaluate their understanding of the topic. Active discussion, as generated by appropriate questions and group activities, are also a crucial opportunity to explore topics outside of the standard curriculum, allowing the diverse experiences of the students to enrich the overall educational experience.

With my background in chemical engineering, I am comfortable teaching any of the core undergraduate curricula in this area. Due to my research interests, I feel I can provide a particularly enriching experience in the areas of thermodynamics, numerical methods, and probabilistic or mathematical methods. In terms of graduate curricula, I envision teaching core topics in thermodynamics, statistical mechanics, and mathematical methods. I am most interested in developing elective courses in molecular simulation methods, as well as practical machine learning techniques. My inspiration for the course material on molecular simulations will be based on my research experiences, as well as recent participation in efforts of the Molecular Sciences Software Institute to detail best-practices in molecular dynamics simulations. This puts me in a unique position to not only teach the subject from the perspective of providing a fundamental understanding of the theory, but also to pass along a dedication to performing research in a reproducible fashion that is at par with the state of the art. Developing a course teaching machine learning from a numerical methods perspective is also particularly exciting to me. Most engineering students have experience with classical numerical methods, but little experience with more modern techniques, which are rapidly growing in popularity and breadth of application. Teaching machine learning from an engineering point of view will be broadly beneficial to graduate and undergraduate chemical engineering students alike. My approach will be to introduce machine learning techniques as “advanced numerical methods,” introducing mathematics as necessary for in-depth understanding, while equally focusing on hands-on exercises based on examples from chemical engineering.

Selected Publications (13 total, 9 first or co-first author, and 1 first-author publication submitted)

Monroe, J. I.; Jiao, S.; Davis, J. R.; Robinson Brown, D.; Katz, L. E.; Shell, M. S. Proceedings of the National Academy of Sciences 2021, 118 (1), e2020205118.

Monroe, J. I.; Hatch, H. W.; Mahynski, N. A.; Shell, M. S.; Shen, V. K. J. Chem. Phys. 2020, 153, 144101.

Monroe, J.; Barry, M.; DeStefano, A.; Gokturk, P. A.; Jiao, S.; Robinson-Brown, D.; Webber, T.; Crumlin, E. J.; Han, S.; Shell, M. S. Annu. Rev. Chem. Biomol. Eng. 2020, 11 (1), annurev-chembioeng-120919-114657.

Monroe, J. I.; Shell, M. S. J. Chem. Phys. 2019, 151 (9), 094501.

Monroe, J. I.; Shell, M. S. Proceedings of the National Academy of Sciences 2018, 115 (32), 8093–8098.

Stock, P.*; Monroe, J. I.*; Utzig, T.; Smith, D. J.; Shell, M. S.; Valtiner, M. ACS Nano 2017, 11 (3). *Co-first authors

Submitted

Monroe, J. I.; Shen, V. K. 2021, Submitted to the Journal of Chemical Physics.

Selected Awards

NRC Postdoctoral Fellowship (2019)

Graduate Opportunity Fellowship, University of California, Santa Barbara (2018)

Graduate Student Award, AICHE COMSEF (2017)

NSF Graduate Research Fellowship (2015)