Yitian Xiao

Hallucination-based AI4Protein design · protein and antibody engineering · computational biology

yitianxi@andrew.cmu.edu · GitHub · WeChat: AIDD小白随想录

I work on hallucination-based AI4Protein design: using structure prediction models as differentiable scoring functions to search sequence space for new antibodies and binders, and following those designs into the wet lab. Most recently I built XDCRCraft for de novo antibody design at XtalPi and ComplexDiff for protein-glue design at Shanghai AI Lab. I am starting an M.S. in Computational Biology at Carnegie Mellon, after a B.S. in Bioinformatics at Zhejiang University.

Experience

1. Apr 2025 – Apr 2026
   XtalPi AI Drug Discovery Algorithm Development Intern, AI4Protein IDD Shanghai

   • Built XDCRCraft, a BindCraft-based de novo antibody hallucination model that integrates an AbLang2 antibody-naturalness prior; roughly doubled the average sequence pass rate.
   • Deployed XDCRCraft across commercial antibody programs for Eli Lilly, AbbVie and Johnson & Johnson, delivering de novo sequences against five targets with successful expression and binding.
   • Led internal antibody design programs end to end, from computational design and screening through sequencing submission for wet-lab validation; advanced four candidates to mouse in vivo studies.
   • Reproduced and optimized SAbDab-based TAP/TNP developability metrics on internal data and integrated the tool into the CADD sequence triage workflow.
2. Apr 2026 – Aug 2026
   Shanghai AI Lab Research Intern, Artificial Intelligence Research Department Shanghai

   • Developed ComplexDiff, a dual-target protein-glue design framework with target-relative SE(3) geometric conditioning and MaSIF-derived interface residues represented as fixed point clouds.
   • Designed the framework for three scenarios: small-molecule glues, redesign of existing complexes, and construction of ternary complexes from pairs of binary complexes.

Education

1. Expected Jun 2028
   Carnegie Mellon University M.S., Computational Biology
2. Jun 2025
   Zhejiang University B.S., Bioinformatics GPA 3.81 / 4.00

Selected publications

Xiao, Y., Zhang, H., Long, Y. (2026). XDCRCraft: Antibody CDR Redesign Method and System Based on Protein Structure Prediction Model Hallucination and Antibody Language Model Regularization. Under review.

Gu, Y., Chen, J., Wang, Z., Shao, Q., Li, Z., Ye, Y., Xiao, Y., et al. (2025). Integrated analysis and systematic characterization of the regulatory network for human germline development. Journal of Genetics and Genomics, 52(2), 204–219. doi:10.1016/j.jgg.2024.11.005

Research projects

1. Feb 2024 – Aug 2024
   AI de novo design and validation of novel antibodies Project lead, Zhejiang University AI Lab, advised by Prof. Ying Chi

   • Built an end-to-end antibody generation and validation pipeline on a RoseTTAFold-based ProteinGenerator; cross-validated sequence and structure pairs with DMFold, ESMFold and AlphaFold 3.
   • Developed a multi-stage ranking scheme across structural plausibility, antigen-binding geometry, molecular-dynamics stability and MM/PBSA binding energy to prioritize the top ten candidates.
2. Mar 2023 – Aug 2023
   Biological study of alpha-olefin bioremediation iGEM modeling team lead

   • Built a SimBiology model of a synthetic-biology system converting alpha-olefins into non-toxic aromatic compounds, used to locate product-release bottlenecks and optimize strain ratios.
3. Jan 2022 – Jan 2023
   Integrated analysis of human germline development Dry-lab researcher, ZJU Stem Cell and Regenerative Medicine Lab

   • Reconstructed developmental regulatory networks with R/Python single-cell RNA-seq analyses; used CellChat and differential-expression analysis to identify cross-stage signaling pathways and key genes.

Toolbox

Programming and ML
PythonPyTorchJAX RLinuxSLURM generative modelingdeep learning

Protein design
de novo protein designantibody and binder design protein language modelsstructure prediction developability screeningmolecular dynamics MM/PBSA

Tools
BindCraftAbLang2AlphaFold 3 ColabFoldESMFoldRoseTTAFold ProteinGeneratorRosetta / PyRosetta

Wet lab
cell culturePCR Western blotimmunoprecipitation

Honors

• National Gold Award, China International College Students' "Internet+" Innovation Competition Biomedicine team lead, 2022
• iGEM Silver Medal Modeling team lead, 2023
• Zhejiang University Second-Class Scholarship 2024
• Zhejiang University Outstanding Graduate 2025

What I write about

Papers and models in AI4Protein: hallucination and diffusion approaches to binder and antibody design, structure prediction and what its confidence scores really mean, design loops that close in the lab, and the parts that still do not work. I write in Chinese on WeChat and keep the English versions here, figures included. Start with the post list; the source of this site is on GitHub.

Corrections are welcome: open an issue on the repository, or email me. The Chinese originals are linked at the end of every post and remain the authoritative version if wording differs.