Odin Zhang[张昊天]


Dual B.S. in Physics and Pharmaceutics at ZJU

Ph.D. in Computer Science at UW Allen School

odinz [AT] uw.edu


Hello! I have earned a dual B.S. in Physics and Pharmaceutics at Zhejiang University in 2022, advised by Prof. Tingjun Hou, focusing on Computer-aided Drug Design After that I am involved in a Master program at Zhejiang University and got an internship in Carbon Silicon AI, a startup company, continuing my research in AI-aided Drug Design. In the future, I will go to UW Allen School for my Ph.D. study, advised by Prof. David Baker for ligand drug design. My CV could be found here.

My research interests are computer-aided drug design. Specifically, I am currently working on (1) structure-based drug design methodology (molecular generation and virtual screening approaches) (2) generative modeling (3) molecular dynamics. Basically, I’m also an AI4Science researcher.


Publications [Google Scholar]

* indicates equal contribution.

Molecular Generation

Structure-based Molecular Generation: A New Paradigm Enables Efficient Design of Potent CLIP1-LTK Inhibitors

Peichen Pan*, ShiCheng Chen*, Odin Zhang*, ..., Tingjun Hou

Cell, Submitted

We extend the Delete model to Casual-Delete model, scussfully design a potent inhibitors targeting to LTK

Delete: Deep Lead Optimization Enveloped in Protein Pocket through unified Deleting Strategies and a Structure-aware Network

Odin Zhang*, Huifeng Zhao*, Xujun Zhang*, ..., Chang-yu Hsieh, Tingjun Hou.

Nat. Commun. Second-round Peer Review

When you face some problems in drug discovery, just delete!

SurfGen: Learning on Topological Surface and Geometric Structure for 3D Molecular Generation

Odin Zhang*, Tianyue Wang*, Gaoqi Weng, ..., Chang-yu Hsieh, Tingjun Hou.

Nat. Compt. Sci, 3, 849–859 (2023)

Inspired from key-and-lock model, we build a 3D molecular design model directly on topological protein surface through Geodesic-GNN and Geoatton-GNN, which satisfy the physical equivariance.

ResGen: Learning the Interaction Patterns through Multiscale Modeling and SE(3)-Framework for 3D-Pocket Molecular Generation

Odin Zhang, Jintu Zhang, Xujun Zhang, ..., Chang-yu Hsieh, Tingjun Hou.

Nat. Mach. Intell., 5, 1020–1030 (2023)

We construct a novel 3D Pocket-aware molecular generation model via multi-scale modeling, achieving SOTA performance on CrossDock dataset and our curated real-world dataset.

Multi-Objective Structure-Based Molecule Generation with Pareto MCTS

Yaodong Yang*, Guangyong Chen, Jinpeng Li, Odin Zhang*, ..., Pheng-Ann Heng


A MCTS-based method for molecular generation

Functional-Group-Based Diffusion for Pocket-Specific Molecule Generation and Elaboration

Haitao Lin, Yufei Huang, Odin Zhang, ..., Stan Z. Li.


We use diffusion model to grow fragments and link fragments for lead optimization

DiffBP: Generative Diffusion of 3D Molecules for Target Protein Binding

Haitao Lin*, Yufei Huang*, Liu Meng, Odin Zhang, Xuanjing Li, Shuiwang Ji, Tingjun Hou, Stan Z. Li

We propose a one-shot approach to capture global interatomic forces for growing molecules directly within protein pockets

A Flexible Data-Free Framework for Structure-Based De Novo Drug Design with Reinforcement Learning

Hongyan Du*, Dejun Jiang*, Odin Zhang, ..., TingjunHou.

Chemical Science

A MCTS-based approach enables de novo design according to the chemicts' requirement

FFLOM: a flow-based autoregressive model for fragment-to-lead optimization. Journal of Medicinal Chemistry

Jieyu Jin, Dong Wang, Odin Zhang, ..., Yu Kang, Tingjun Hou

Journal of Medicinal Chemistry 14(8), 2054-2069

We develop a flow-based model for fragment elaboration and linker design, achieving the SOTA performance regarding 2D metrics

Other AIDD approaches

KarmaDock: a deep learning paradigm for ultra-large library docking with fast speed and high accuracy

Xujun Zhang*, Odin Zhang*, ..., Chang-yu Hsieh, Tingjun Hou.

Nat. Compt. Sci.,3, 739–740 (2023).

A Deep Learning-based docking method, from binding pose to virtual screening

Highly accurate and efficient deep learning paradigm for full-atom protein loop modeling with KarmaLoop

Tianyue Wang*, Xujun Zhang*, Odin Zhang*, ..., Chang-yu Hsieh, Tingjun Hou.

Nat. Mach. Intell., Major Revision

A SOTA method for the geometry prediction of the loop domain in proteins

SDEGen: Learning to Evolve Molecular Conformations from Thermodynamic Noise for Conformation Generation

Odin Zhang*, ShengMing Li*, Jintu Zhang, .., Yu Kang, Tingjun Hou.

Chem. Sci., 14, 1557-1568

We construct a novel conformation generation model via the stochastic differential equation-based generative model, which is a kind of diffusion process.

Protein 3D Graph Structure Learning for Robust Structure-based Protein Property Prediction

Yufei Huang*, ..., Odin Zhang, ..., Stan. Z. Li.

AAAI 2024

In this paper, we first investigate the rea- son behind the decrease in prediction accuracy when utilizing predicted structures, attributing it to the structural distribution bias from the perspective of structure representation learning

Infinite Physical Monkey: Do Deep Learning Methods Really Perform Better in Conformation Generation?

Odin Zhang, Jintu Zhang, Huifeng Zhao, Dejun Jiang, Yafeng Deng.

Bio Arxiv

We present our opinions on conformation generation, discussing another two papers on arxiv platform

MetalProGNet: A Structure-based Deep Graph Model for Metalloprotein–ligand Interaction Predictions

Dejun Jiang, Zhaofeng Ye, Chang-yu Hsieh, Odin Zhang, ..., Jian Wu, Tingjun Hou

Chem. Sci. 14(8), 2054-2069

We develop a novel and accurate scoring function for metalloproteins, the main methodology is bi-partite graph and GNN

Explainable Fragment-Based Molecular Property Attribution

Linxiang Jia, Zunlei Feng, Odin Zhang, Jie Song, Zipeng Zhong, Shaolun Yao, Mingli Song.

Adv. Intell. Syst. 4(10), 2200104

I led the experimental validation of the model throught collection of pharmacological data, this is my early-stage participation in undergraduate period

Classical CADD and Molecular Dynamics

Quasiclassical Trajectory Simulation as a Protocol to Build Locally Accurate Machine Learning Potentials

Jintu Zhang*, Odin Zhang*, Zhixin Qin, Yu Kang, Xin Hong, Tingjun Hou.

J. Chem. Inf. Model. 2023, 63, 4, 1133–1142

We show that quasiclassical trajectory (QCT) calculations can be invoked to efficiently obtain locally accurate ML-PESs, thus accelearing molecular dynamics

An Engineered Design of Self-Assembly Nanomedicine Guided by Molecular Dynamic Simulation for Photodynamic and Hypoxia-Directed Therapy

Yu Wang, Odin Zhang, Jianwei Wang, Guping Tang, Hongzhen Bai.

Mol. Pharmaceutics, 2023, 20, 4, 2128–2137

I conduct the MD and DPD to guide the design of the nanomedicine ISDNN, which achieves uniform size distribution and high drug loading up to 90%

Sigmoid Accelerated Molecular Dynamics: An Efficient Enhanced Sampling Method for Biosystems

Yihao Zhao, Jintu Zhang, Odin Zhang, Shukai Gu, Yafeng Deng, Yaoquan Tu, Tingjun Hou, Yu Kang.

J. Phys. Chem. Lett. 2023, 14, 4, 1103–1112

We propose sigmoid accelerated molecular dynamics (SaMD), which can produce the free energy landscape with better accuracy and efficiency compared with GaMD.

Amphiphilic porphyrin-based supramolecular self-assembly for photochemotherapy: From molecular design to application

Shuping Wang, Xuechun Huang, Yunhong He, Odin Zhang*, Jun Zhou, Guping Tang, Shijun Li, Hongzhen Bai.

Nano Today, 48, 101732

I conduct the Disspative Particle Dynamics (DPD) to study the stability of CPT@P complex, this is my early-stage work in undergraduate period

How Good Are Current Docking Programs at Nucleic acids-ligand Docking? a Comprehensive Evaluation

Dejun Jiang, Huifeng Zhao, ...,Odin Zhang, Tingjun Hou

Journal of Chemical Theory and Computation 14(8), 2054-2069

We conducted a comprehensive evaluation of molecular docking programs for NA–ligand systems, contributing to the future research about virtual screening on such a system


Research_Blog could be found in my Github

RDKit Blog, contains the dihedral finding, conformation generation, even electronic density computation!

PDB Blog, contains the use of PDBParser, you could find the cripts for downloading AlphaFold-Predicted Structures, PDBfile->fasta format, and even sequence alignment!

Matplotlib Blog, a informative figure make it easier for readers to understand research data.


Dedicate to HanShan and The Dharma Bums

It's cold in the mountains, and not just this year.
The jagged cliffs are always piled high with snow, and the deep canyon species of trees are always spitting mist outward.
The grass is still sprouting in June, and the leaves start falling in August when they first arrive.
And here I was, mesmerized like a drug addict who had just finished a hit.

*A conversation at the bottom of the heart*
You have fulfilled your long-standing dream, and broken through all obstacles.
Why do you still feel so painful?

At the moment I realise my dream,
I really felt great satisfaction at the bottom of my heart.
But almost in an instant,
I felt the endless nothingness.
What did the aurora bring to me?
There seems to be nothing.
I still clearly remember the palpitation in my heart when I dreamed of seeing the aurora as a child,
but when I grew up, the throbbing became lighter and lighter.
I was overwhelmed by life, and was no longer able to complete the so-called dream.
But when I buried my dream,
my heart was unwilling.
So, I' ve tried my best,
and came here with the body that I didn't know how long I could live,
I long for the aurora to bring me an eternal meaning,
so that I can fight (against) life bravely.
But that doesn't exist.
It has no meaning at all.
My life will not change, and my story will not fly to the blue sky.
In that case, what is the meaning of life?
The dream that can be moved for it will be dull after it is realized.
All our pursues will turn into nothingness in the end.
Why should we purse it?
So, I'm not going to go on.
Since life has no meaning, let me go.
Let me go to the abyss, to embrace the eternal nothingness,
to embrace the eternal death.

Don't lie to yourself.
Deep In your heart, you're not here to see the aurora.
You just want to escape.
In order to escape your present, past and future life.
But after escaping?
You still have to come back and face everything.
You don't just want to give up because life is meaningless,
you're afraid,
fear of how you will spend your painful and hopeless life in the future.
In the past, you put your hopes for the future on something that seems meaningful and valuable.
-watching the aurora.
But that's just a mirage.
When you really realize this, you surrender to life
-since watching the aurora is meaningless,
since life is meaningless,
why continue to live?
It is because the world has no meaning in the present, in the past or in the future,
and because life has no meaning in the first place,
that we are freed from the illusion, and are able to be real with the world of nothingness.
Leave meaning behind, it doesn't matter.
It is only after all the old ideas have been broken down and reorganised that we can be reborn.
to go back to the roots of everything,
to do something in order to do it.
The meaning of life is...
Go to the dock and order some French fries!

I don’t wanta hear all your word descriptions of words words words you made up all winter, man I wanta be enlightened by actions.