(364i) Bridging Physics-Based Simulations and AI-Driven Methods to Accelerate the Design of the Next-Generation Polymers

Polymers, ubiquitous in modern life, play a critical role in various sectors such as textiles, automotive, construction, and healthcare, collectively contributing to a market exceeding $600 billion annually. As the world grapples with escalating challenges like climate change, plastic waste, food security and public health crises, there emerges a pressing demand for pioneering next-generation polymeric materials. However, innovations in polymeric materials lag due to the complexities and disparate behaviors arising from various parameters in monomer structures, compositions, and topological structures, as well as the knowledge gap between microstructural features and macroscopic responses and applicability. My research interests will focus on bridging physics-based simulations and AI-driven approaches to accelerate the understanding and development of advanced polymeric materials for functionality, sustainability, healthcare, and biotechnology. I will investigate the intricate behavior and properties of polymers, identify key performance determinants, and facilitate the strategic inverse design of novel polymer materials by integrating machine learning, molecular simulations, enhanced sampling, chemical informatics, optimization and language models.

Research Experience

Advancing polymer similarity calculation and property predictions through machine learning and informatics, Department of Chemical Engineering, Massachusetts Institute of Technology (advised by Bradley D. Olsen and Debra J. Audus (NIST))

My postdoctoral research focuses on developing polymer similarity and embedding by using machine learning and chemical informatics for building polymer database platforms and advancing polymer properties prediction. Defining the similarity between chemical entities is an essential task in polymer informatics, enabling ranking, clustering, and classification. Despite its importance, the pairwise chemical similarity of polymers remains an open problem. I developed chemically intuitive functions that utilize the earth mover’s distance and graph edit distance to accurately quantify the pairwise similarity. These functions effectively capture the chemical, topological, and ensemble features of polymers, representing a significant advancement toward the creation of robust polymer database search engines and quantitative polymer design tools. Additionally, I have designed and implemented a graph neural network model, MacroSimGNN, which accelerates the calculation of pairwise macromolecular similarity while maintaining high levels of accuracy and interpretability. To further enhance the utility of polymer informatics, I introduced a landmark distance embedding based on polymer similarity algorithms. This new embedding method has proven effective in accurately predicting the physical properties of complex polymers and offers a novel approach to handling the diversity of polymer structures. My contributions extend to fostering an open benchmark that accelerates the standardization and collaborative, open-source development within this research area. I am also actively engaged in several interdisciplinary collaborations that apply advanced AI techniques to challenges in chemistry and materials science, where I utilized prompt engineering with large language models to automatically identify relevant literature, facilitating the transformation of unstructured scientific texts into structured datasets.

Computing free energy landscapes for materials design, Department of Chemical and Biomolecular Engineering, University of Notre Dame (advised by Jonathan K. Whitmer)

My Ph.D. research focused on investigating the free energy surfaces of soft materials and advancing the inverse design of materials through the integration of machine learning, molecular simulations, and advanced sampling algorithms. Understanding polymer-surface interactions is crucial for numerous industrial and biomedical applications, yet the quantitative analysis and design of polymer sequences for these interactions remains an unresolved challenge. To address this, I integrated coarse-grained molecular simulations with data-driven machine learning methods to predict polymer-surface adhesive free energy, significantly accelerating the optimal design of functional polymers. In tackling the issue of data scarcity, a common obstacle in computational materials science, I employed transfer learning to significantly enhance prediction performance. This method proved invaluable in contexts where material data is limited, demonstrating its potential to transform materials property prediction and design practices. Additionally, my research included using molecular simulations coupled with enhanced sampling methods to study the temperature dependence of materials' physical properties. The insights gained from the study of the elastic response of liquid crystals were particularly impactful in guiding the design of advanced liquid crystal sensors. The investigation of the isomerization kinetics of clusters provides a detailed quantitative understanding of these materials' dynamic properties at finite temperatures.

Awards:

• Future Faculty Scholar, ACS Polymeric Materials Science and Engineering (PMSE) (2023)
• Big Data Award at Big Data in Polymer Chemistry Session, ACS POLY (2023)
• Finalist, NIST Postdoctoral & Early-career Association of Researchers (PEAR) Accolades Outstanding Technical (2023)
• Winner, MIT ChemE Teach-Off Competition (2023)
• Travel Grant of Forum for Early Career Scientists (FECS), APS March (2023)
• Travel Grant of Division of Soft Matter (DSOFT), APS March (2020)
• Outstanding Paper Award, Department of Chemical and Biomolecular Engineering, University of Notre Dame (2020)
• Best Poster Award, 6th Annual Notre Dame-Purdue Soft Matter & Polymers Symposium (2019)

Publications:

[1] Jiale Shi, Debra J. Audus, Bradley D. Olsen. “MacroSimGNN: Graph Neural Network for Efficient Macromolecular Similarity Calculation.” In preparation.

[2] Jiale Shi, Katharina A. Fransen, Nathan J. Rebello, Debra J. Audus, Bradley D. Olsen. “OpenBenchmark for Polymer Informatics.” In preparation.

[3] Katharina A. Fransen, Julia Casey, Jiale Shi, Natalie D. Mamrol, Gabrielle F. Godbille-Cardona, Jiarui Lu, Alex M. Zappi, Debra J. Audus, Bradley D. Olsen. “Predictive Modeling for Polyester Biodegradability: Insights from Augmenting Binary Co-polyester Data with a Terpolymer Library.” In Preparation.

[4] Nathan J. Rebello, Akash Arora, Hidenobu Mochigase, Tzyy-Shyang Lin, Jiale Shi, Debra J. Audus, Eric. S. Muckley, and Bradley D. Olsen. “BCDB: The Block Copolymer Phase Behavior Database.” Journal of Chemical Information and Modeling Accepted.

[5] Joren Van Herck,.., Jiale Shi,..., Jonathan Whitmer,..., Berend Smit. “Assessment of Fine-Tuned Large Language Models for Real-World Chemistry and Material Science Applications.” Journal of the American Chemical Society submitted.

[6] Qianxiang Ai, Fanwang Meng, Jiale Shi, Brenden Pelkie, Connor W. Coley. “Extracting Structured Data from Organic Synthesis Procedures Using a Fine-Tuned Large Language Model.” ChemRxiv 2024.

[7] Jiale Shi, Dylan Walsh, Nathan J. Rebello, Weizhong Zou, Michael E. Deagen, Katharina A. Fransen, Xian Gao, Bradley D. Olsen, Debra J. Audus. “Calculating Pairwise Similarity of Polymer Ensembles via Earth Mover’s Distance.” ACS Polymers Au 2024, 4, 1, 66–76.

[8] Jiale Shi, Nathan J. Rebello, Dylan Walsh, Weizhong Zou, Michael Deagen, Bruno Salomao Leao, Debra J. Audus, Bradley D. Olsen. “Quantifying Pairwise Similarity for Complex Polymers.” Macromolecules 2023, 56, 18, 7344-7357.

[9] Kevin Maik Jablonka,..., Jiale Shi,...“14 Examples of How LLMs Can Transform Materials Science and Chemistry: A Reflection on a Large Language Model Hackathon.” Digital Discovery 2023, 2, 1233-1250.

[10] Jiale Shi, Fahed Albreiki, Yamil J. Colón, Samanvaya Srivastava, Jonathan K. Whitmer. “Using Transfer Learning to Leverage Prior Knowledge in the Prediction of Adhesive Free Energies between Polymers and Surfaces.” Journal of Chemical Theory and Computation. 2023, 19, 14, 4631-4640.

[11] Jiale Shi, Michael J. Quevillon, Pedro Henrique Amorim Valença, Jonathan K. Whitmer. “Predicting Adhesive Free Energies of Polymer-Surface Interactions with Machine Learning.” ACS Applied Materials & Interfaces 2022, 14, 32, 37161–37169.

[12] Jiale Shi, Shanghui Huang, François Gygi, Jonathan K. Whitmer. “Free Energy Landscape and Isomerization Rates of Au4 Clusters at Finite Temperature.” The Journal of Physical Chemistry A 2022, 126, 21, 3392-3400.

[13] Jiale Shi*, Hythem Sidky*, Jonathan K. Whitmer. “Automated determination of n-cyanobiphenyl and n-cyanobiphenyl binary mixtures elastic constants in the nematic phase from molecular simulation.” Molecular Systems Design & Engineering 2020, 5, 1131-1136. (* indicates equal contribution and co-first authorship)

[14] Jiale Shi, Hythem Sidky, Jonathan K. Whitmer. “Novel Elastic Response in Twist-bend Nematic Models.” Soft Matter 2019, 15, 8219-8226. (inside front cover)