Ziwei Yang, Ph.D.

Projects

Research software, computational methods, and AI systems developed through my research.

01
Model Biology

Mechanistic Biological Modeling

GraphGDel

IEEE TCBB 2026

Builds graph representations of genome-scale metabolic models and learns from them, together with gene and metabolite sequences, to predict growth-coupled gene deletion strategies.

GraphGDel architecture detail: the Graph-M graph encoder over the metabolic graph and the Pred-M predictor

DeepGDel

IEEE TCBB 2025

A deep learning framework that learns gene and metabolite sequence representations to predict gene deletion strategies for growth-coupled production.

DeepGDel Meta-M module: an encoder-decoder that learns latent representations of metabolite sequences

DBgDel

IEEE TCBB 2025

Mines gene deletion databases for prior knowledge to narrow the search space, computing growth-coupled gene deletion strategies 6.1× faster on average than existing methods.

DBgDel step 1: shared remaining genes are computed from known deletion strategies in the MetNetComp database
02
Understand Biology

Multimodal Biomedical AI

GeSubNet

ICLR 2025 · Oral

Generates disease-subtype-specific gene interaction networks by combining a generative model of patient gene expression with a graph neural network over prior gene networks.

A general breast cancer gene network with unknown links to patient subtypes Luminal A, Luminal B and Basal-like
03
Discover Biology

AI Agents for Scientific Discovery

Biomedical AI Copilots

Making Large Language Models Reliable Data Science Programming Copilots for Biomedical Research

Nature Biomedical Engineering 2026

Benchmarks LLMs on 293 biomedical data science coding tasks, where overall accuracy stays below 40%, and shows that an agent which plans and refines its analysis before coding reaches 74%.

BioDSBench contents: coding tasks, analysis results, test cases and an execution sandbox