CV

Basics

Name Hongwei Tu
Email hongweit at andrew dot cmu dot edu
Github https://github.com/LiamT01

Education

  • 2023.08 - 2024.12

    Pittsburgh, PA, USA

    Master of Science
    Carnegie Mellon University
    Computer Science
  • 2019.09 - 2023.06

    Shanghai, China

    Bachelor of Engineering
    Shanghai Jiao Tong University
    Artificial Intelligence

Projects

  • 2023.10 - 2024.11
    Learning Interactions of Intrinsically Disordered Proteins (IDPs) with DNA
    First Author | Advisor: Prof. Jian Ma | Ma Lab, CMU
    • Developed the first deep learning approach to explore protein-DNA interactions for multiple IDPs.
    • Achieved 0.84 Pearson correlation in efficiently predicting 20 kb base-resolution binding profiles using U-Net.
    • Generalized across cell types with accuracy comparable to experimental reproducibility.
    • Identified IDP-binding motifs and their spatial distribution, cooperative interactions, and binding preferences.
    • Submitted a paper to RECOMB 2025.
  • 2022.07 - 2023.05
    Accelerating Molecular Dynamics Simulations of RNAs with a Graph Neural Network
    Bachelor’s Thesis | Advisor: Prof. Jinjin Li | AIMS Lab, SJTU
    • Developed a graph neural network (GNN) to predict molecular force fields for RNAs, achieving 8% relative loss.
    • Integrated the GNN with a simulation tool to replace potential energy surface models for >1000× speedups.
    • Improved the GNN's generalization ability through unsupervised data augmentation.
  • 2022.05 - 2023.09
    Predicting Molecular Properties with a Rotationally Invariant Graph Neural Network
    First Author | Advisor: Prof. Jinjin Li | AIMS Lab, SJTU
    • Designed a graph neural network (RotNet) with a rotationally invariant transformation, overcoming the generalization deficiency caused by rotations of molecules.
    • Generalized well to various molecular conformations and achieved 2% relative loss, outperforming prior state-of-the-art methods.
  • 2021.10 - 2022.09
    Predicting Protein Energy Changes upon Single-Point Amino Acid Mutations
    First Author | Advisor: Prof. Jinjin Li | AIMS Lab, SJTU
    • Designed the Clustered Tree Regression (CTR) method, integrating unsupervised and supervised learning.
    • Achieved a 6% improvement in accuracy compared to previous methods, using protein sequences alone as input.
    • Developed software for CTR with copyright license No. 2022SR1040914 granted in China.

Work

  • 2024.08 - 2024.12
    Teaching Assistant
    Machine Learning Department, Carnegie Mellon University
    10-701 Introduction to Machine Learning (PhD)
    • Created homework assignments and exams, hosted recitations and office hours, and mentored projects.
  • 2024.01 - 2024.05
    Teaching Assistant
    Machine Learning Department, Carnegie Mellon University
    10-701 Introduction to Machine Learning (PhD)
    • Created homework assignments and exams, hosted recitations and office hours, and mentored projects.

Publications

Awards