Xuzheng Tian

Advisor: Prof. Karl Jacob

will propose a doctoral thesis entitled,

Machine-Learning-Assisted Modeling of Polymeric Nanoparticles and Their Multicomponent Interactions 

On

Monday, December 8th at 12:30 pm

 Virtually via Zoom

Link

Committee

            Prof. Karl Jacob– School of Materials Science and Engineering

            Prof. Donggang Yao– School of Materials Science and Engineering

            Prof. Youjiang Wang– School of Materials Science and Engineering

            Prof. Hamid Garmestani – School of Materials Science and Engineering

            Prof. Edmond Chow – School of Computational Science and Engineering

Abstract

Polymeric nanoparticle composites play an essential role in multiple fields such as drug delivery and advanced composite design, yet their performance is governed by complex multicomponent interactions that are difficult to predict using traditional modeling approaches. Molecular simulations provide valuable atomistic insight but remain limited by computational cost and scale, while existing empirical models fail to generalize across diverse polymer-drug and polymer-inorganic interfaces. In this proposal, I present an AI-enhanced framework that integrates experiment and simulation data with graph-based machine learning to study polymeric nanoparticles across multiple length scales. In the first part of this thesis, I developed a Message Passing Neural Network-Transformer model that incorporates chemical structure, copolymer composition, and cross-attention mechanisms to predict long-acting injectable (LAI) drug release kinetics. By combining empirical kinetic modeling with learned structural representations, the approach offers improved predictive accuracy and interpretability. Building on this foundation, the second part will implement this Transformer framework into a Generative model to predict polymers structures based on certain interaction with respect with the target component. And in the third part, I will focus on understanding interface confinement in polymer-nanofiller composites using MD-assisted learning and AI model to explore the inorganic and organic materials' interaction. Together, these studies aim to establish a comprehensive data-driven framework for predicting and designing multicomponent polymeric systems.