Seonghwan Seo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Seonghwan Seo has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Woo Youn Kim
Department of Chemistry, KAIST, Daejeon, Republic of Korea
5 shared publications - Jaechang Lim
HITS Inc., Seoul, Republic of Korea
3 shared publications Benedikt M. Wanner
Molecules Inc., Kemptthal, Switzerland
1 shared publicationDaniil Melnichenko
HITS Inc., Seoul, Republic of Korea
1 shared publication- Hyeongwoo Kim
Department of Chemistry, KAIST, Daejeon, Republic of Korea
1 shared publication Jihye Choi
HITS Inc., Seoul, Republic of Korea
1 shared publication
Latest Publications
- Open Access
Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery
Autonomous drug discovery via generative molecular design is critically bottlenecked by the production of chemically intractable structures. Multi-trillion-scale make-on-demand libraries guarantee synthetic feasibility, but conventional virtual screening cannot efficiently navigate these vast spaces or address multi-parameter optimization. We introduce Hyper Screening X, a generative AI framework integrating structure-based design with automated synthesis hardware. By encoding deterministic reaction logic in a generative flow network, the framework evaluates only 10 million candidates to navigate an 11-trillion-compound space, jointly optimizing physicochemical properties while ensuring compatibility with automated synthesis. Prospective validation against an SLC1A5 variant with a cryptic interface yielded a 96% synthesis success rate and a 50% functional hit rate, identifying two first-in-class lead compounds. This synthesis-aware strategy advances autonomous, closed-loop drug discovery.
Open AccessSynthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery
Autonomous drug discovery via generative molecular design is critically bottlenecked by the production of chemically intractable structures. Multi-trillion-scale make-on-demand libraries guarantee synthetic feasibility, but conventional virtual screening cannot efficiently navigate these vast spaces or address multi-parameter optimization. We introduce Hyper Screening X, a generative AI framework integrating structure-based design with automated synthesis hardware. By encoding deterministic reaction logic in a generative flow network, the framework evaluates only 10 million candidates to navigate an 11-trillion-compound space, jointly optimizing physicochemical properties while ensuring compatibility with automated synthesis. Prospective validation against an SLC1A5 variant with a cryptic interface yielded a 96% synthesis success rate and a 50% functional hit rate, identifying two first-in-class lead compounds. This synthesis-aware strategy advances autonomous, closed-loop drug discovery.
- Open Access
AtlasFold: Protein structure prediction with metagenomic-scale language models
Protein language models (PLMs) trained on evolutionary sequences learn representations that encode protein structure, enabling direct structure prediction without multiple-sequence alignments (MSAs). Here we present the Atlas model family, an open and trainable system spanning protein language modeling, monomer folding, and protein-complex prediction. AtlasLM-3B is a 3B-scale language model trained with masked language modeling on approximately 1.56 billion sequences, including metagenomic data, and outperforms the similarly sized ESM2-3B in unsupervised contact prediction. Building on these representations, AtlasFold predicts all-atom protein structures and achieves state-of-the-art accuracy among PLM-based folding models. Fine-tuning AtlasFold for protein-complex prediction produces AtlasFold-Multimer (AtlasFold-M), whose antibody–antigen prediction performance is comparable to that of AlphaFold3 and ESMFold2. This protein-specific folding architecture enables fast, memory-efficient inference with AtlasFold and AtlasFold-M. By releasing the training code and data, stage checkpoints, and model weights under the MIT License, we provide a foundation for advancing PLM-based protein structure prediction. 1
Open AccessAtlasFold: Protein structure prediction with metagenomic-scale language models
Protein language models (PLMs) trained on evolutionary sequences learn representations that encode protein structure, enabling direct structure prediction without multiple-sequence alignments (MSAs). Here we present the Atlas model family, an open and trainable system spanning protein language modeling, monomer folding, and protein-complex prediction. AtlasLM-3B is a 3B-scale language model trained with masked language modeling on approximately 1.56 billion sequences, including metagenomic data, and outperforms the similarly sized ESM2-3B in unsupervised contact prediction. Building on these representations, AtlasFold predicts all-atom protein structures and achieves state-of-the-art accuracy among PLM-based folding models. Fine-tuning AtlasFold for protein-complex prediction produces AtlasFold-Multimer (AtlasFold-M), whose antibody–antigen prediction performance is comparable to that of AlphaFold3 and ESMFold2. This protein-specific folding architecture enables fast, memory-efficient inference with AtlasFold and AtlasFold-M. By releasing the training code and data, stage checkpoints, and model weights under the MIT License, we provide a foundation for advancing PLM-based protein structure prediction. 1
- Open Access
- 4 November 2024
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 15 (46) , 19473-19487
PharmacoNet: deep learning-guided pharmacophore modeling for ultra-large-scale virtual screening
PharmacoNet is developed for virtual screening, including deep learning-guided protein-based pharmacophore modeling, a parameterized analytical scoring function, and coarse-grained pose alignment. It is extremely fast yet reasonably accurate.
- 4 November 2024
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 15 (46) , 19473-19487
Open AccessPharmacoNet: deep learning-guided pharmacophore modeling for ultra-large-scale virtual screening
PharmacoNet is developed for virtual screening, including deep learning-guided protein-based pharmacophore modeling, a parameterized analytical scoring function, and coarse-grained pose alignment. It is extremely fast yet reasonably accurate.
- Open Access
- 3 January 2023
- Published by Wiley in Advanced Science
- Vol. 10 (8) , e2206674
Molecular Generative Model via Retrosynthetically Prepared Chemical Building Block Assembly
Deep generative models are attracting attention as a smart molecular design strategy. However, previous models often render molecules with low synthesizability, hindering their real‐world applications. Here, a novel graph‐based conditional generative model which makes molecules by tailoring retrosynthetically prepared chemical building blocks until achieving target properties in an auto‐regressive fashion is proposed. This strategy improves the synthesizability and property control of the resulting molecules and also helps learn how to select appropriate building blocks and bind them together to achieve target properties. By applying a negative sampling method to the selection process of building blocks, this model overcame a critical limitation of previous fragment‐based models, which can only use molecules from the training set during generation. As a result, the model works equally well with unseen building blocks without sacrificing computational efficiency. It is demonstrated that the model can generate potential inhibitors with high docking scores against the 3CL protease of SARS‐COV‐2.
- 3 January 2023
- Published by Wiley in Advanced Science
- Vol. 10 (8) , e2206674
Open AccessMolecular Generative Model via Retrosynthetically Prepared Chemical Building Block Assembly
Deep generative models are attracting attention as a smart molecular design strategy. However, previous models often render molecules with low synthesizability, hindering their real‐world applications. Here, a novel graph‐based conditional generative model which makes molecules by tailoring retrosynthetically prepared chemical building blocks until achieving target properties in an auto‐regressive fashion is proposed. This strategy improves the synthesizability and property control of the resulting molecules and also helps learn how to select appropriate building blocks and bind them together to achieve target properties. By applying a negative sampling method to the selection process of building blocks, this model overcame a critical limitation of previous fragment‐based models, which can only use molecules from the training set during generation. As a result, the model works equally well with unseen building blocks without sacrificing computational efficiency. It is demonstrated that the model can generate potential inhibitors with high docking scores against the 3CL protease of SARS‐COV‐2.
- Open Access
- 14 December 2021
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 13 (2) , 554-565
Drug-likeness scoring based on unsupervised learning
A new quantification method of drug-likeness based on unsupervised learning. The method only uses drug molecules as training set without any non-drug-like molecules.
- 14 December 2021
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 13 (2) , 554-565
Open AccessDrug-likeness scoring based on unsupervised learning
A new quantification method of drug-likeness based on unsupervised learning. The method only uses drug molecules as training set without any non-drug-like molecules.