COMA: efficient structure-constrained molecular generation using contractive and margin losses

Journal of Cheminformatics

Research

Sanghyun Park¹  Sangmin Seo²  Jonghwan Choi² 

[1] Department of Computer Science, Yonsei University, Yonsei-ro 50, 03722, Seoul, Republic of Korea;Department of Computer Science, Yonsei University, Yonsei-ro 50, 03722, Seoul, Republic of Korea;UBLBio Corporation, Yeongtong-ro 237, 16679, Suwon, Gyeonggi-do, Republic of Korea;

关键词: Drug design;    Molecular optimization;    Goal-directed molecular generation;    Structure-constrained molecular generation;    Deep generative model;    Metric learning;    Contrastive learning;    Reinforcement learning;   

DOI  :  10.1186/s13321-023-00679-y

 received in 2022-10-05, accepted in 2023-01-04,  发布年份 2023

来源: Springer

【 摘 要 】

BackgroundStructure-constrained molecular generation is a promising approach to drug discovery. The goal of structure-constrained molecular generation is to produce a novel molecule that is similar to a given source molecule (e.g. hit molecules) but has enhanced chemical properties (for lead optimization). Many structure-constrained molecular generation models with superior performance in improving chemical properties have been proposed; however, they still have difficulty producing many novel molecules that satisfy both the high structural similarities to each source molecule and improved molecular properties.MethodsWe propose a structure-constrained molecular generation model that utilizes contractive and margin loss terms to simultaneously achieve property improvement and high structural similarity. The proposed model has two training phases; a generator first learns molecular representation vectors using metric learning with contractive and margin losses and then explores optimized molecular structure for target property improvement via reinforcement learning.ResultsWe demonstrate the superiority of our proposed method by comparing it with various state-of-the-art baselines and through ablation studies. Furthermore, we demonstrate the use of our method in drug discovery using an example of sorafenib-like molecular generation in patients with drug resistance.

【 授权许可】

CC BY   
© The Author(s) 2023

【 图 表 】

41116_2022_35_Article_IEq123.gif

41116_2022_35_Article_IEq122.gif

Fig. 2

42004_2022_796_Article_IEq4.gif

41116_2022_35_Article_IEq118.gif

41116_2022_35_Article_IEq117.gif

41116_2022_35_Article_IEq116.gif

41116_2022_35_Article_IEq115.gif

Fig. 16

Fig. 1

41116_2022_35_Article_IEq114.gif

41116_2022_35_Article_IEq113.gif

41116_2022_35_Article_IEq112.gif

41116_2022_35_Article_IEq111.gif

Fig. 4

Fig. 13

41116_2022_35_Article_IEq110.gif

41116_2022_35_Article_IEq109.gif

41116_2022_35_Article_IEq108.gif

41116_2022_35_Article_IEq106.gif

41116_2022_35_Article_IEq105.gif

41116_2022_35_Article_IEq104.gif

Fig. 3

Fig. 1

41116_2022_35_Article_IEq102.gif

Fig. 6

Fig. 3

41116_2022_35_Article_IEq101.gif

41116_2022_35_Article_IEq100.gif

41235_2023_464_Article_IEq27.gif

Fig. 6

41116_2022_35_Article_IEq97.gif

41116_2022_35_Article_IEq95.gif

41116_2022_35_Article_IEq94.gif

Fig. 6

41116_2022_35_Article_IEq93.gif

41116_2022_35_Article_IEq92.gif

41116_2022_35_Article_IEq90.gif

Fig. 4

Fig. 1

41116_2022_35_Article_IEq87.gif

Fig. 4

40854_2022_413_Article_IEq262.gif

Fig. 3

Fig. 2

41116_2022_35_Article_IEq84.gif

41116_2022_35_Article_IEq83.gif

Fig. 1

Fig. 2

Fig. 4

41116_2022_35_Article_IEq81.gif

41116_2022_35_Article_IEq80.gif

41116_2022_35_Article_IEq78.gif

Fig. 6

41116_2022_35_Article_IEq77.gif

41116_2022_35_Article_IEq76.gif

41116_2022_35_Article_IEq75.gif

41116_2022_35_Article_IEq74.gif

Fig. 1

Fig. 12

40249_2022_1049_Article_IEq39.gif

Fig. 9

Fig. 8

40249_2022_1049_Article_IEq33.gif

Fig. 4

Fig. 5

41116_2022_35_Article_IEq47.gif

Fig. 6

41116_2022_35_Article_IEq40.gif

41116_2022_35_Article_IEq38.gif

41116_2022_35_Article_IEq37.gif

40249_2022_1049_Article_IEq26.gif

Fig. 3

41116_2022_35_Article_IEq35.gif

Fig. 4

41116_2022_35_Article_IEq33.gif

41116_2022_35_Article_IEq29.gif

41116_2022_35_Article_IEq28.gif

Fig. 1

Fig. 1

12936_2022_4438_Article_IEq29.gif

Fig. 1

Fig. 3

Fig. 1

Fig. 1

41116_2022_35_Article_IEq11.gif

Fig. 4

Fig. 3