PUBLICATIONS
2025
4.
박준영; 유선용
Abstract | Links | BibTeX | Dimensions | Tags: Bioinformatics, Drugs, Transformer
@article{박준영;유선용2025,
title = {화합물의 골격구조를 활용한 Transformer 기반 새로운 분자 설계},
author = {박준영 and 유선용},
url = {http://journal.dcs.or.kr/_common/do.php?a=full&b=12&bidx=3950&aidx=43776},
doi = {10.9728/dcs.2025.26.1.217},
issn = {1598-2009},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {디지털콘텐츠학회논문지},
volume = {26},
number = {1},
pages = {217-223},
abstract = {전통적인 신약 개발은 새로운 약물을 시장에 출시하기까지 많은 시간과 막대한 비용이 소요되며, 높은 실패율로 인해 효율성이 낮다는 문제가 있다. 이러한 문제를 해결하기 위해 생성 모델을 활용한 혁신적인 접근법이 주목받고 있다. 본 연구에서는 트랜스포머 디코더 구조를 기반으로 화합물의 구조 정보를 문자열로 학습하여 새로운 화합물 구조를 생성하는 모델을 제안한다. 특히, 화합물에서 추출한 골격 구조(scaffold)를 임베딩하여 모델 입력에 포함함으로써, 결합 및 원자 정보와 골격 구조를 동시에 처리하였다. 벤치마크 데이터셋을 사용한 평가 결과, 골격 구조 임베딩을 적용한 모델이 데이터셋 별로 유효성 지표에서 0.964, 0.986의 우수한 성능을 보였다. 본 연구는 분자 생성 모델에 골격 구조 임베딩을 도입함으로써, 화학적 규칙을 준수하는 분자를 효과적으로 생성할 수 있는 방법을 제시하였으며, 신약 개발 분야에서 AI 기반 분자 설계의 효율성을 높이는 데 기여할 것으로 기대된다.},
keywords = {Bioinformatics, Drugs, Transformer},
pubstate = {published},
tppubtype = {article}
}
전통적인 신약 개발은 새로운 약물을 시장에 출시하기까지 많은 시간과 막대한 비용이 소요되며, 높은 실패율로 인해 효율성이 낮다는 문제가 있다. 이러한 문제를 해결하기 위해 생성 모델을 활용한 혁신적인 접근법이 주목받고 있다. 본 연구에서는 트랜스포머 디코더 구조를 기반으로 화합물의 구조 정보를 문자열로 학습하여 새로운 화합물 구조를 생성하는 모델을 제안한다. 특히, 화합물에서 추출한 골격 구조(scaffold)를 임베딩하여 모델 입력에 포함함으로써, 결합 및 원자 정보와 골격 구조를 동시에 처리하였다. 벤치마크 데이터셋을 사용한 평가 결과, 골격 구조 임베딩을 적용한 모델이 데이터셋 별로 유효성 지표에서 0.964, 0.986의 우수한 성능을 보였다. 본 연구는 분자 생성 모델에 골격 구조 임베딩을 도입함으로써, 화학적 규칙을 준수하는 분자를 효과적으로 생성할 수 있는 방법을 제시하였으며, 신약 개발 분야에서 AI 기반 분자 설계의 효율성을 높이는 데 기여할 것으로 기대된다.
2024
3.
Yeabean Na; Junho Kim; Myung-Gyun Kang; Sunyong Yoo
Abstract | Links | BibTeX | Tags: Bioinformatics, Deep learning, Drugs
@conference{Yoo2024,
title = {A Multimodal Deep Learning Approach for Predicting Drug Metabolism According to the CYP2D6 Genetic Variation},
author = {Yeabean Na and Junho Kim and Myung-Gyun Kang and Sunyong Yoo},
url = {https://dtmbio.net/},
year = {2024},
date = {2024-01-02},
urldate = {2024-01-02},
publisher = {The 18th International Conference on Data and Text Mining in Biomedical Informatics},
abstract = {Background Cytochrome P450 2D6 (CYP2D6) is involved in metabolizing up to 25% of the drugs commonly used in clinics. Characterized by high polymorphisms, CYP2D6 is one of the key pharmacogenes in pharmacogenomics. This genetic variability can lead to significant inter-patient differences in drug metabolism, resulting in differential therapeutic responses and adverse effects. However, conducting in vivo or in vitro experiments for each CYP2D6 variant across various drugs is time-consuming, ethically challenging, and expensive. Given these constraints, In silico modeling approaches for predicting the drug metabolism profiles of CYP2D6 variants are a critical necessity.
Methods A multimodal deep learning approach that combined CYP2D6 genotype data and drug structural information was used in this study. A Convolutional Neural Network (CNN) was used to encode the genotype data, and a Graph Convolutional Network (GCN) was used to decode the drug structures. These diverse data types were then integrated into a multimodal model to predict drug metabolism.
Results A comparative analysis was conducted between a CNN model utilizing solely the CYP2D6 genotype data and a multimodal model incorporating both genotype and drug-specific information. The multimodal approach demonstrated better performance across all evaluated metrics. An additional experiment predicting drug metabolism on unseen drug data also performed well.
Conclusions This model is anticipated to enhance the prediction of metabolic capacity in previously uncharacterized CYP2D6 variants, potentially reducing adverse drug reactions.},
keywords = {Bioinformatics, Deep learning, Drugs},
pubstate = {published},
tppubtype = {conference}
}
Background Cytochrome P450 2D6 (CYP2D6) is involved in metabolizing up to 25% of the drugs commonly used in clinics. Characterized by high polymorphisms, CYP2D6 is one of the key pharmacogenes in pharmacogenomics. This genetic variability can lead to significant inter-patient differences in drug metabolism, resulting in differential therapeutic responses and adverse effects. However, conducting in vivo or in vitro experiments for each CYP2D6 variant across various drugs is time-consuming, ethically challenging, and expensive. Given these constraints, In silico modeling approaches for predicting the drug metabolism profiles of CYP2D6 variants are a critical necessity.
Methods A multimodal deep learning approach that combined CYP2D6 genotype data and drug structural information was used in this study. A Convolutional Neural Network (CNN) was used to encode the genotype data, and a Graph Convolutional Network (GCN) was used to decode the drug structures. These diverse data types were then integrated into a multimodal model to predict drug metabolism.
Results A comparative analysis was conducted between a CNN model utilizing solely the CYP2D6 genotype data and a multimodal model incorporating both genotype and drug-specific information. The multimodal approach demonstrated better performance across all evaluated metrics. An additional experiment predicting drug metabolism on unseen drug data also performed well.
Conclusions This model is anticipated to enhance the prediction of metabolic capacity in previously uncharacterized CYP2D6 variants, potentially reducing adverse drug reactions.
Methods A multimodal deep learning approach that combined CYP2D6 genotype data and drug structural information was used in this study. A Convolutional Neural Network (CNN) was used to encode the genotype data, and a Graph Convolutional Network (GCN) was used to decode the drug structures. These diverse data types were then integrated into a multimodal model to predict drug metabolism.
Results A comparative analysis was conducted between a CNN model utilizing solely the CYP2D6 genotype data and a multimodal model incorporating both genotype and drug-specific information. The multimodal approach demonstrated better performance across all evaluated metrics. An additional experiment predicting drug metabolism on unseen drug data also performed well.
Conclusions This model is anticipated to enhance the prediction of metabolic capacity in previously uncharacterized CYP2D6 variants, potentially reducing adverse drug reactions.
2018
2.
Sunyong Yoo; Suhyun Ha; Moonshik Shin; Kyungrin Noh; Hojung Nam; Doheon Lee
Abstract | Links | BibTeX | Dimensions | Tags: Database, Drugs, Ethnopharmacology, Machine learning
@article{yoo2018data,
title = {A data-driven approach for identifying medicinal combinations of natural products},
author = {Sunyong Yoo and Suhyun Ha and Moonshik Shin and Kyungrin Noh and Hojung Nam and Doheon Lee},
url = {https://ieeexplore.ieee.org/abstract/document/8482294},
doi = {10.1109/ACCESS.2018.2874089},
year = {2018},
date = {2018-10-05},
urldate = {2018-10-05},
journal = {IEEE Access},
volume = {6},
pages = {58106–58118},
publisher = {IEEE},
abstract = {Combinations of natural products have been used as important sources of disease treatments. Existing databases contain information about prescriptions, herbs, and compounds and their relationships with phenotypes, but they do not have information on the use of combinations of natural product compounds. In this paper, we identified large-scale associations between natural product combinations and phenotypes by applying an association rule mining technique to integrated information on herbal medicine, combination drugs, functional foods, molecular compounds, and target genes. The rationale behind this approach is that natural products commonly found in medicinal multicomponent mixtures have statistically significant associations with the therapeutic effects of the multicomponent mixtures. Based on a molecular network analysis and an external literature validation, we show that the inferred associations are valuable information for identifying medicinal combinations of natural products since they have statistically significant closeness proximity in the molecular layer and have much experimental evidence. All results are available through the workbench site at http://biosoft.kaist.ac.kr/coconut to facilitate the investigation of the medicinal use of natural products and their combinations.},
keywords = {Database, Drugs, Ethnopharmacology, Machine learning},
pubstate = {published},
tppubtype = {article}
}
Combinations of natural products have been used as important sources of disease treatments. Existing databases contain information about prescriptions, herbs, and compounds and their relationships with phenotypes, but they do not have information on the use of combinations of natural product compounds. In this paper, we identified large-scale associations between natural product combinations and phenotypes by applying an association rule mining technique to integrated information on herbal medicine, combination drugs, functional foods, molecular compounds, and target genes. The rationale behind this approach is that natural products commonly found in medicinal multicomponent mixtures have statistically significant associations with the therapeutic effects of the multicomponent mixtures. Based on a molecular network analysis and an external literature validation, we show that the inferred associations are valuable information for identifying medicinal combinations of natural products since they have statistically significant closeness proximity in the molecular layer and have much experimental evidence. All results are available through the workbench site at http://biosoft.kaist.ac.kr/coconut to facilitate the investigation of the medicinal use of natural products and their combinations.
1.
Sunyong Yoo; Kyungrin Noh; Moonshik Shin; Junseok Park; Kwang-Hyung Lee; Hojung Nam; Doheon Lee
Abstract | Links | BibTeX | Dimensions | Tags: ADR, Bioinformatics, Drugs, Network analysis
@article{yoo2018silico,
title = {In silico profiling of systemic effects of drugs to predict unexpected interactions},
author = {Sunyong Yoo and Kyungrin Noh and Moonshik Shin and Junseok Park and Kwang-Hyung Lee and Hojung Nam and Doheon Lee},
url = {https://www.nature.com/articles/s41598-018-19614-5},
doi = {10.1038/s41598-018-19614-5},
year = {2018},
date = {2018-01-25},
urldate = {2018-01-25},
journal = {Scientific Reports},
volume = {8},
number = {1},
pages = {1612},
publisher = {Nature Publishing Group UK London},
abstract = {Identifying unexpected drug interactions is an essential step in drug development. Most studies focus on predicting whether a drug pair interacts or is effective on a certain disease without considering the mechanism of action (MoA). Here, we introduce a novel method to infer effects and interactions of drug pairs with MoA based on the profiling of systemic effects of drugs. By investigating propagated drug effects from the molecular and phenotypic networks, we constructed profiles of 5,441 approved and investigational drugs for 3,833 phenotypes. Our analysis indicates that highly connected phenotypes between drug profiles represent the potential effects of drug pairs and the drug pairs with strong potential effects are more likely to interact. When applied to drug interactions with verified effects, both therapeutic and adverse effects have been successfully identified with high specificity and sensitivity. Finally, tracing drug interactions in molecular and phenotypic networks allows us to understand the MoA.},
keywords = {ADR, Bioinformatics, Drugs, Network analysis},
pubstate = {published},
tppubtype = {article}
}
Identifying unexpected drug interactions is an essential step in drug development. Most studies focus on predicting whether a drug pair interacts or is effective on a certain disease without considering the mechanism of action (MoA). Here, we introduce a novel method to infer effects and interactions of drug pairs with MoA based on the profiling of systemic effects of drugs. By investigating propagated drug effects from the molecular and phenotypic networks, we constructed profiles of 5,441 approved and investigational drugs for 3,833 phenotypes. Our analysis indicates that highly connected phenotypes between drug profiles represent the potential effects of drug pairs and the drug pairs with strong potential effects are more likely to interact. When applied to drug interactions with verified effects, both therapeutic and adverse effects have been successfully identified with high specificity and sensitivity. Finally, tracing drug interactions in molecular and phenotypic networks allows us to understand the MoA.