danДоклады Национальной академии наук БеларусиDoklady of the National Academy of Sciences of Belarus1561-83232524-2431The Republican Unitary Enterprise Publishing House "Belaruskaya Navuka"10.29235/1561-8323-2025-69-1-13-22dan-1231Research ArticleИНФОРМАТИКАINFORMATICSГенеративная состязательная нейронная сеть на основе графовых эмбеддингов для de novo дизайна низкомолекулярных ингибиторов фермента KasA микобактерии туберкулезаGenerative adversarial neural network with graph embeddings for de novo designing small-molecule inhibitors against Mycobacterium tuberculosis KasA enzymeГончарА. В.GoncharA. V.

Гончар Анна Викторовна – мл. науч. сотрудник

ул. Сурганова, 6, 220012, Минск

Gonchar Anna V. – Junior Researcher

6, Surganov Str., 220012, Minsk

hanna.hanchar@gmail.comФурсК. В.FursK. V.

Фурс Константин Викторович – мл. науч. сотрудник

ул. Сурганова, 6, 220012, Минск

Furs Konstantin V. – Junior Researcher

6, Surganov Str., 220012, Minsk

kvfurs@gmail.comТузиковА. В.TuzikovA. V.

Тузиков Александр Васильевич – член-корреспондент, д-р физ.-мат. наук, профессор, заведующий лабораторией

ул. Сурганова, 6, 220012, Минск

Tuzikov Alexander V. – Corresponding Member, D. Sc. (Physics and Mathematics), Professor, Head of the Laboratory

6, Surganov Str., 220012, Minsk

tuzikov@newman.bas-net.byАндриановА.AndrianovA. M.

Андрианов Александр Михайлович – д-р хим. наук, профессор, гл. науч. сотрудник

ул. Купревича, 5/2, 220084, Минск

Andrianov Alexander M. – D. Sc. (Chemistry), Professor, Chief Researcher

5/2, Kuprevich Str., 220084, Minsk

alexande.andriano@yandex.ruОбъединенный институт проблем информатики Национальной академии наук БеларусиUnited Institute of Informatics Problems of the National Academy of Sciences of BelarusИнститут биоорганической химии Национальной академии наук БеларусиInstitute of Bioorganic Chemistry of the National Academy of Sciences of Belarus2025260220256911322Copyright © Гончар А.В., Фурс К.В., Тузиков А.В., Андрианов А.M., 20252025Гончар А.В., Фурс К.В., Тузиков А.В., Андрианов А.Gonchar A.V., Furs K.V., Tuzikov A.V., Andrianov A.M.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://doklady.belnauka.by/jour/article/view/1231

Разработана генеративная состязательная нейронная сеть с частичным привлечением учителя, обученная на графовых эмбеддингах и предназначенная для de novo дизайна потенциальных ингибиторов бета-кето- ацил-[ацил-белок-носитель] синтазы I (KasA) - фермента, критически важного для биосинтеза миколовых кислот клеточной стенки микобактерии туберкулеза. Проведено обучение и тестирование созданной нейронной сети на наборе соединений из виртуальной библиотеки малых молекул, содержащих элементы структуры, способные к селективным взаимодействиям с терапевтической мишенью. С помощью разработанной нейронной сети осуществлен de novo дизайн 3637 соединений с последующей оценкой потенциала их ингибиторной активности против белка KasA методами молекулярного докинга. На основе анализа полученных данных отобраны шесть соединений, проявляющих высокое сродство к малонил-связывающему сайту фермента. Предполагается, что идентифицированные соединения формируют перспективные базовые структуры для проведения дальнейших теоретических и экспериментальных исследований по разработке новых эффективных ингибиторов лекарственно-устойчивых форм туберкулеза.

A generative semi-supervised adversarial neural network trained on graph embeddings was developed for de novo design of potential inhibitors against beta-ketoacyl-[acyl-carrier protein] synthase I (KasA), an enzyme critically important for biosynthesis of mycolic acids of the Mycobacterium tuberculosis cell wall. The designed model was trained and tested on a set of compounds from a virtual library of small molecules containing structural elements capable of selective interactions with the therapeutic target. Using the developed neural network, 3,637 compounds were de novo designed, followed by assessment of their inhibitory activity against the KasA protein using molecular docking methods. Based on the analysis of the obtained data, six compounds exhibiting high affinity to the malonyl-binding site of the enzyme were selected. The identified compounds are assumed to form promising basic structures for further theoretical and experimental studies on the development of new effective inhibitors of drug-resistant tuberculosis.

микобактерия туберкулезафермент KasAгенеративная состязательная нейронная сетьобучение с частичным привлечением учителяграфовые эмбеддингивиртуальный скринингмолекулярный докингMycobacterium tuberculosisKasA enzymegenerative adversarial neural networksemi-supervised learninggraph embeddingsvirtual screeningmolecular dockingReferencesA comprehensive survey of prospective structure-based virtual screening for early drug discovery in the past fifteen years / H. Zhu, Y. Zhang, W. Li, N. Huang // International Journal of Molecular Sciences. – 2022. – Vol. 23, N 24. – Art. 15961. https://doi.org/10.3390/ijms232415961A comprehensive survey of prospective structure-based virtual screening for early drug discovery in the past fifteen years / H. Zhu, Y. Zhang, W. Li, N. Huang // International Journal of Molecular Sciences. – 2022. – Vol. 23, N 24. – Art. 15961. https://doi.org/10.3390/ijms232415961Virtual screening algorithms in drug discovery: A review focused on machine and deep learning methods / T. A. D. Oli veira, M. P. D. Silva, E. H. B. Maia [et al.] // Drugs and Drug Candidates. – 2023. – Vol. 2, N 2. – P. 311–334. https://doi.org/10.3390/ddc2020017Virtual screening algorithms in drug discovery: A review focused on machine and deep learning methods / T. A. D. Oli veira, M. P. D. Silva, E. H. B. Maia [et al.] // Drugs and Drug Candidates. – 2023. – Vol. 2, N 2. – P. 311–334. https://doi.org/10.3390/ddc2020017Identification of new Mycobacterium tuberculosis proteasome inhibitors using a knowledge-based computational screening approach / T. M. Almeleebia, M. A. Shahrani, M. Y. Alshahrani [et al.] // Molecules. – 2021. – Vol. 26, N 8. – Art. 2326. https://doi.org/10.3390/molecules26082326Identification of new Mycobacterium tuberculosis proteasome inhibitors using a knowledge-based computational screening approach / T. M. Almeleebia, M. A. Shahrani, M. Y. Alshahrani [et al.] // Molecules. – 2021. – Vol. 26, N 8. – Art. 2326. https://doi.org/10.3390/molecules26082326A deep learning approach to antibiotic discovery / J. M. Stokes, K. Yang, K. Swanson [et al.] // Cell. – 2020. – Vol. 180, N 4. – P. 688–702. https://doi.org/10.1016/j.cell.2020.01.021A deep learning approach to antibiotic discovery / J. M. Stokes, K. Yang, K. Swanson [et al.] // Cell. – 2020. – Vol. 180, N 4. – P. 688–702. https://doi.org/10.1016/j.cell.2020.01.021Drug discovery for Mycobacterium tuberculosis using structure-based computer-aided drug design approach / M. A. Ejalonibu, S. A. Ogundare, A. A. Elrashedy [et al.] // International Journal of Molecular Sciences. – 2021. – Vol. 22, N 24. – Art. 13259. https://doi.org/10.3390/ijms222413259Drug discovery for Mycobacterium tuberculosis using structure-based computer-aided drug design approach / M. A. Ejalonibu, S. A. Ogundare, A. A. Elrashedy [et al.] // International Journal of Molecular Sciences. – 2021. – Vol. 22, N 24. – Art. 13259. https://doi.org/10.3390/ijms222413259Identification of KasA as the cellular target of an anti-tubercular scaffold / K. A. Abrahams, C. W. Chung, S. Ghidelli-Disse [et al.] // Nature Communications. – 2016. – Vol. 7. – Art. 12581. https://doi.org/10.1038/ncomms12581Identification of KasA as the cellular target of an anti-tubercular scaffold / K. A. Abrahams, C. W. Chung, S. Ghidelli-Disse [et al.] // Nature Communications. – 2016. – Vol. 7. – Art. 12581. https://doi.org/10.1038/ncomms12581Conditional depletion of KasA, a key enzyme of mycolic acid biosynthesis, leads to mycobacterial cell lysis / A. Bhatt, L. Kremer, A. Z. Dai [et al.] // Journal of Bacteriology. – 2005. – Vol. 187, N 22. – P. 7596–7606. https://doi.org/10.1128/jb.187.22.7596-7606.2005Conditional depletion of KasA, a key enzyme of mycolic acid biosynthesis, leads to mycobacterial cell lysis / A. Bhatt, L. Kremer, A. Z. Dai [et al.] // Journal of Bacteriology. – 2005. – Vol. 187, N 22. – P. 7596–7606. https://doi.org/10.1128/jb.187.22.7596-7606.2005Odena, A. Semi-supervised learning with generative adversarial networks / A. Odena. – 2016. – URL: https://arxiv.org/pdf/1606.01583 (date of access: 13.01.2025).Odena, A. Semi-supervised learning with generative adversarial networks / A. Odena. – 2016. – URL: https://arxiv.org/pdf/1606.01583 (date of access: 13.01.2025).Jin, W. Junction tree variational autoencoder for molecular graph generation / W. Jin, R. Barzilay, T. Jaakkola // Inter national conference on machine learning. – 2018. – P. 2323–2332.Jin, W. Junction tree variational autoencoder for molecular graph generation / W. Jin, R. Barzilay, T. Jaakkola // Inter national conference on machine learning. – 2018. – P. 2323–2332.Weininger, D. SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules / D. Weininger // Journal of Chemical Information and Computer Sciences. – 1988. – Vol. 28, N 1. – P. 31–36. https://doi.org/10.1021/ci00057a005Weininger, D. SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules / D. Weininger // Journal of Chemical Information and Computer Sciences. – 1988. – Vol. 28, N 1. – P. 31–36. https://doi.org/10.1021/ci00057a005Slow onset inhibition of bacterial beta-ketoacyl-acyl carrier protein synthases by thiolactomycin / C. A. Machutta, G. R. Bommineni, S. R. Luckner [et al.] // Journal of Biological Chemistry. – 2010. – Vol. 285, N 9. – P. 6161–6169. https://doi.org/10.1074/jbc.m109.077909Slow onset inhibition of bacterial beta-ketoacyl-acyl carrier protein synthases by thiolactomycin / C. A. Machutta, G. R. Bommineni, S. R. Luckner [et al.] // Journal of Biological Chemistry. – 2010. – Vol. 285, N 9. – P. 6161–6169. https://doi.org/10.1074/jbc.m109.077909Conversion of a beta-ketoacyl synthase to a malonyl decarboxylase by replacement of the active-site cysteine with glutamine / A. Witkowski, A. K. Joshi, Y. Lindqvist, S. Smith // Biochemistry. – 1999. – Vol. 38, N 36. – P. 11643–11650. https://doi.org/10.1021/bi990993hConversion of a beta-ketoacyl synthase to a malonyl decarboxylase by replacement of the active-site cysteine with glutamine / A. Witkowski, A. K. Joshi, Y. Lindqvist, S. Smith // Biochemistry. – 1999. – Vol. 38, N 36. – P. 11643–11650. https://doi.org/10.1021/bi990993hStructural basis for the recognition of mycolic acid precursors by KasA, a condensing enzyme and drug target from Mycobacterium tuberculosis / J. Schiebel, K. Kapilashrami, A. Fekete [et al.] // Journal of Biological Chemistry. – 2013. – Vol. 288, N 47. – P. 34190–34204. https://doi.org/10.1074/jbc.m113.511436Structural basis for the recognition of mycolic acid precursors by KasA, a condensing enzyme and drug target from Mycobacterium tuberculosis / J. Schiebel, K. Kapilashrami, A. Fekete [et al.] // Journal of Biological Chemistry. – 2013. – Vol. 288, N 47. – P. 34190–34204. https://doi.org/10.1074/jbc.m113.511436Exponential consensus ranking improves the outcome in docking and receptor ensemble docking / K. Palacio-Rodriguez, I. Lans, C. N. Cavasotto, P. Cossio // Scientific Reports. – 2019. – Vol. 9. – Art. 5142. https://doi.org/10.1038/s41598-019-41594-3Exponential consensus ranking improves the outcome in docking and receptor ensemble docking / K. Palacio-Rodriguez, I. Lans, C. N. Cavasotto, P. Cossio // Scientific Reports. – 2019. – Vol. 9. – Art. 5142. https://doi.org/10.1038/s41598-019-41594-3

The authors declare that there are no conflicts of interest present.