Seonghwan Seo

AI Researcher for Drug Discovery

Bio

This is Seonghwan Seo, a Ph.D. student in the Department of Chemistry, KAIST. I am currently a member of the Intelligent Chemistry Lab under the supervision of Prof. Woo Youn Kim.
My research area is AI-based material discovery, with a particular focus on small drug molecules. I have developed deep learning models in various areas of drug discovery including pharmacophore modeling, virtual screening, property prediction, and generative models. Looking ahead, my interest will be extended to binding pose and binding affinity prediction.

(Last Updated: 2024.01.04)

Research Experiences

Ph.D. Candidates

KAIST
Feb. 2022 - Present
  • The first deep learning model for fully-automated receptor-based pharmacophore modeling
  • Development of molecular voxelization tool (MolVoxel)
  • Selectivity-aware ligand scoring model for virtual screening

AI Scientist Intern

HITS Inc.
Dec. 2020 - Aug. 2022
  • Building block-based graph generative model for synthesizable molecular design
  • Reaction template-based SMILES generative model for synthesizable molecular design
  • Protein-ligand binding pose prediction by score-based modeling

Undergraduated Researcher

KAIST
June 2019 - Dec. 2021
  • Designing the deep learning framework for a new drug-likeness scoring

Publication

* indicates equal contribution

Journals

Molecular generative model via retrosynthetically prepared chemical building block assembly
Seonghwan Seo, Jaechang Lim, Woo Youn Kim
Advanced Science, 2023
[Paper] [Github]

Drug-likeness scoring based on unsupervised learning
Kyunghoon Lee*, Jinho Jang*, Seonghwan Seo*, Jaechang Lim, Woo Youn Kim
Chemical Science, 2022
[Paper] [Github]

Conferences

PharmacoNet: Accelerating Large-Scale Virtual Screening by Deep Pharmacophore Modeling
Seonghwan Seo, Woo Youn Kim
NeurIPS 2023 Workshop on New Frontiers of AI for Drug Discovery and Development (AI4D3)
[OpenReview] [ArXiv] [Github]

Journal Cover Image

Molecular generative model via retrosynthetically prepared chemical building block assembly
Seonghwan Seo, Jaechang Lim, Woo Youn Kim
Advanced Science, 2023
[Cover]

Softwares

MolVoxel: Easy-to-Use Molecular Voxelization Tool
Seonghwan Seo
[PyPi] [Github]

Research Area

  • Small Molecule
  • Protein
  • Pharmacophore Modeling
  • Generative Model
  • Multi-Modal
  • XAI
  • Graph
  • 3D Image

Education

  • Ph.D. in Chemistry
    Korea Advanced Institute of Science and Technology (KAIST)
    Aug 2022 - Aug 2027 (expected)
  • B.S. in Chemistry & Computer Science (Double Major)
    Korea Advanced Institute of Science and Technology (KAIST)
    GPA: 3.99/4.3
    Feb 2018 - Aug 2022

Skills & Tools

Deep Learning

  • Python
  • PyTorch
  • PyTorch Geometric
  • Numba
  • PyTorch Lightning

Chemistry

  • RDKit
  • OpenBabel

Language

  • Korean (Native)
  • English (Upper-intermediate)