2026

Protein Deimmunization with a Generative Diffusion Model
Protein Deimmunization with a Generative Diffusion Model

Zijie Chen*, Chih-Chieh (Jay) Yu*, Cai Zhou, Xinyun Jiang, Long T. Nguyen, Cameron Myhrvold, Tommi Jaakkola, Kaiyi Jiang (* equal contribution)

bioRxiv 2026

MESSI redesigns protein sequences to reduce predicted MHC-II presentation risk while preserving structure and function. Recursive lookahead-feedback guides structure-conditioned discrete diffusion, with functional redesigns validated experimentally in luciferase, anti-PD-1 minibinders, and uricases.

Protein Deimmunization with a Generative Diffusion Model
Protein Deimmunization with a Generative Diffusion Model

Zijie Chen*, Chih-Chieh (Jay) Yu*, Cai Zhou, Xinyun Jiang, Long T. Nguyen, Cameron Myhrvold, Tommi Jaakkola, Kaiyi Jiang (* equal contribution)

bioRxiv 2026

MESSI redesigns protein sequences to reduce predicted MHC-II presentation risk while preserving structure and function. Recursive lookahead-feedback guides structure-conditioned discrete diffusion, with functional redesigns validated experimentally in luciferase, anti-PD-1 minibinders, and uricases.

Rethinking Diffusion Models with Symmetries through Canonicalization with Applications to Molecular Graph Generation
Rethinking Diffusion Models with Symmetries through Canonicalization with Applications to Molecular Graph Generation

Cai Zhou*, Zijie Chen*, Zian Li, Jike Wang, Kaiyi Jiang, Pan Li, Rose Yu, Muhan Zhang, Stephen Bates, Tommi Jaakkola (* equal contribution)

ICML GenBio Workshop 2026

A canonicalization framework for symmetry-invariant generative modeling that removes symmetry-induced ambiguity in diffusion training and improves 3D molecule generation.

Rethinking Diffusion Models with Symmetries through Canonicalization with Applications to Molecular Graph Generation

Cai Zhou*, Zijie Chen*, Zian Li, Jike Wang, Kaiyi Jiang, Pan Li, Rose Yu, Muhan Zhang, Stephen Bates, Tommi Jaakkola (* equal contribution)

ICML GenBio Workshop 2026

A canonicalization framework for symmetry-invariant generative modeling that removes symmetry-induced ambiguity in diffusion training and improves 3D molecule generation.

Revisiting Target-Aware de novo Molecular Generation with TarPass: Between Rational Design and Texas Sharpshooter
Revisiting Target-Aware de novo Molecular Generation with TarPass: Between Rational Design and Texas Sharpshooter

Rui Qin*, Zijie Chen*, Yurong Li, Meijing Fang, Longji Shen, Yilong Su, Odin Zhang, Qinghan Wang, Qun Su, Jike Wang$^{\dagger}$, Tingjun Hou$^{\dagger}$, Yu Kang$^{\dagger}$ (* equal contribution)

Advanced Science 2026

TarPass, a rigorous benchmark for target-aware de novo molecular generation, building a scalable evaluation pipeline and uncovering a critical gap between favorable docking metrics and faithful protein–ligand interaction modeling.

Revisiting Target-Aware de novo Molecular Generation with TarPass: Between Rational Design and Texas Sharpshooter

Rui Qin*, Zijie Chen*, Yurong Li, Meijing Fang, Longji Shen, Yilong Su, Odin Zhang, Qinghan Wang, Qun Su, Jike Wang$^{\dagger}$, Tingjun Hou$^{\dagger}$, Yu Kang$^{\dagger}$ (* equal contribution)

Advanced Science 2026

TarPass, a rigorous benchmark for target-aware de novo molecular generation, building a scalable evaluation pipeline and uncovering a critical gap between favorable docking metrics and faithful protein–ligand interaction modeling.