vedomostiregmedРегуляторные исследования и экспертиза лекарственных средствRegulatory Research and Medicine Evaluation3034-30623034-3453Federal State Budgetary Institution ‘Scientific Centre for Expert Evaluation of Medicinal Products’ of the Ministry of Health of the Russian Federation (FSBI ‘SCEEMP’)10.30895/1991-2919-2025-15-6-630-641vedomostiregmed-817Research ArticleЦИФРОВЫЕ ТЕХНОЛОГИИ В ФАРМАЦИИDIGITAL TECHNOLOGIES AND PHARMACYНейросетевые технологии в регистрации лекарственных средств: методология машинного анализа документов и интеллектуальных систем реального времениNeural Network Technologies in Drug Registration: Computerised Analysis of Documents and Real-Time Systemshttps://orcid.org/0009-0008-5302-7609ЯрошинскийМ. А.YaroshinskyM. A.

Ярошинский Милан Анатольевич

Верейская ул, д. 29, стр. 134, Москва, 121357

Milan A. Yaroshinsky

29/134 Vereyskaya St., Moscow 121357

myaroshinskiy@pharm-sintez.ruhttps://orcid.org/0009-0008-9805-2402АндрееваМ. В.AndreevaM. V.

Андреева Мария Владимировна

Верейская ул, д. 29, стр. 134, Москва, 121357

Maria V. Andreeva

29/134 Vereyskaya St., Moscow 121357

mandreeva@pharm-sintez.ruhttps://orcid.org/0000-0001-5545-135XБалакинЕ. И.BalakinE. I.

Балакин Евгений Игоревич, канд. мед. наук

Верейская ул, д. 29, стр. 134, Москва, 121357

Evgenii I. Balakin, Cand. Sci. (Med.)

29/134 Vereyskaya St., Moscow 121357

evgbalakin@yandex.ruhttps://orcid.org/0000-0003-2734-5036СавченкоА. Ю.SavchenkoA. Yu.

Савченко Алла Юрьевна, канд. мед. наук

Каширское ш., д. 31, Москва, 115409

Alla Yu. Savchenko, Cand. Sci. (Med.)

31 Kashirskoe Hwy, Moscow 115409

AYSavchenko@mephi.ruhttps://orcid.org/0009-0006-4636-0978ПавловА. С.PavlovA. S.

Павлов Александр Сергеевич

Миусская пл., д. 9, Москва, 125047

Alexander S. Pavlov

9 Miusskaya Sq., Moscow 125047

lup.mail@gmail.comhttps://orcid.org/0000-0003-3396-5813ПустовойтВ. И.PustovoitV. I.

Пустовойт Василий Игоревич, д-р мед. наук

ул. Живописная, д. 46, корп. 8, Москва, 123098

Vasily I. Pustovoit, Dr. Sci. (Med.)

46 Zhivopisnaya St., Moscow 123098

vipust@yandex.ruАкционерное общество «Фарм-Синтез»РоссияPharm-Sintez AORussian FederationФедеральное государственное автономное образовательное учреждение высшего образования «Национальный исследовательский ядерный университет «МИФИ» (НИЯУ МИФИ)РоссияNational Research Nuclear University MEPhI (Moscow Engineering Physics Institute)Russian FederationФедеральное государственное бюджетное учреждение высшего образования «Российский химико-технологический университет им. Д.И. Менделеева»РоссияDmitry Mendeleev University of Chemical Technology of RussiaRussian FederationФедеральное государственное бюджетное учреждение «Государственный научный центр Российской Федерации – Федеральный медицинский биофизический центр имени А.И. Бурназяна» Федерального медико-биологического агентства РоссииРоссияState Research Center – Burnasyan Federal Medical Biophysical Center of Federal Medical Biological AgencyRussian Federation202519122025156630641Copyright © Ярошинский М.А., Андреева М.В., Балакин Е.И., Савченко А.Ю., Павлов А.С., Пустовойт В.И., 20252025Ярошинский М.А., Андреева М.В., Балакин Е.И., Савченко А.Ю., Павлов А.С., Пустовойт В.И.Yaroshinsky M.A., Andreeva M.V., Balakin E.I., Savchenko A.Y., Pavlov A.S., Pustovoit V.I.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://www.vedomostincesmp.ru/jour/article/view/817ВВЕДЕНИЕ

ВВЕДЕНИЕ. Существующие методы подготовки документов, используемые в ходе разработки лекарственных средств, характеризуются высокими временными затратами (40–60% рабочего времени специалистов), высокой частотой ошибок, вносимых в документацию, и ограниченной интероперабельностью данных. Увеличение эффективности подготовки документов возможно при использовании нейросетевых технологий и переходе к комплексной автоматизации процедур жизненного цикла регистрационного досье.

ЦЕЛЬ

ЦЕЛЬ. Оценка возможности использования систем искусственного интеллекта (ИИ) и машинного анализа при подготовке регистрационного досье лекарственного препарата в процессе разработки лекарственного средства.

ОБСУЖДЕНИЕ

ОБСУЖДЕНИЕ. Модели обработки естественного языка (NLP) показывают высокую эффективность в области обработки технической и регуляторной документации. Системы распознавания именованных сущностей (NER) с точностью извлечения 89–96% позволяют сократить время обработки (подготовки и последующей проверки) производителем материалов при формировании электронного общего технического документа на 64%, однако возможность обработки информации ограничена трудностями интерпретации морфологически сложных терминов и требует использования наборов аннотированных данных. Следует отметить, что генеративные модели типа GPT-4 без дополнительной настройки при использовании в архитектуре генерации, дополненной поиском (RAG), могут формировать фактологически некорректную информацию. Предиктивные системы на основе графовых нейросетей и ансамблей XGBoost демонстрируют высокую точность (ROC AUC до 0,88) при прогнозировании одобрения препаратов, но имеют недостаток в виде невозможности интерпретации решений и систематических смещений в данных. Разработка документоцентричных платформ с NLP позволяет сократить время подготовки досье на 60%, однако при внедрении автоматизированной процедуры формирования разделов досье требуется экспертная верификация.

ВЫВОДЫ

ВЫВОДЫ. Концепция интегрированных ИИ-систем подтверждает свою эффективность, сокращая сроки обработки документов производителем и повышая точность решений, что способствует ускорению вывода препаратов на рынок. Перспективы внедрения цифровых технологий связаны с преодолением различий в терминах через унифицированные онтологии. Для практической реализации требуется разработка единых стандартов валидации ИИ-алгоритмов и адаптивных систем.

INTRODUCTION

INTRODUCTION. Current methods of handling medicine regulatory documents are associated with high time cost (40-60% of labour hours), frequent documentation errors, and limited data interoperability. Neural network technologies have enabled the enhanced document preparation and a transition to full automation of the registration dossier life cycle.

AIM

AIM. This study aimed to evaluate the possibility of using artificial intelligence (AI) systems in preparing a drug registration dossier.

DISCUSSION

DISCUSSION. Natural language processing (NLP) models demonstrate high efficiency for the regulatory documentation. Named entity recognition (NER) systems with 89–96% entity extraction accuracy rate reduces the processing (preparation and quality review) time for documents within electronic Common Technical Document (eCTD) by 64%, but face limitations in interpreting morphologically complex terms and require annotated datasets. Without additional fine-tuning, generative models such as GPT-4, are prone to generating inaccurate facts when used in the Retrieval-Augmented Generation (RAG) architecture. Predictive systems based on graph neural networks and XGBoost ensembles demonstrate high accuracy (ROC AUC up to 0.88) when predicting drug approval; however, they cannot interpret decisions and data systematic biases. Developing document-centric platforms with NLP reduces the dossier preparation time by 60%, still, implementing an automated procedure for generating dossier sections requires an expert verification.

CONCLUSIONS

CONCLUSIONS. The concept of integrated AI systems proves its effectiveness by reducing the document handling time by manufacturers and increasing the accuracy of decisions, which in turn speeds up the market launch of medicinal products. The prospects of introducing digital technologies are associated with overcoming definitions differences through unified ontologies. Practical implementation requires the development of unified standards for the validation of AI algorithms and adaptive systems.

 

 

регистрационное досьеискусственный интеллектобработка естественного языкапредиктивное моделированиедокументоцентричные платформыBioBERTстандарты ALCOA+IDAAPM-базавалидация алгоритмовэОТД-досьеобщий технический документregistration dossierartificial intelligencenatural language processingpredictive modellingdocument-centric platformsBioBERTALCOA+ standardsIDAAPM databasealgorithm validationeCTD dossierРаботы выполнена без спонсорской поддержки.The authors declare no conflict of interest.ReferencesNiazi SK. Bridging the regulatory divide: a dual-pathway framework using SRA approvals and AI evaluation to ensure drug quality in developing countries. Pharmaceuticals (Basel). 2025;18(7):1024. https://doi.org/10.3390/ph18071024Niazi SK. Bridging the regulatory divide: a dual-pathway framework using SRA approvals and AI evaluation to ensure drug quality in developing countries. Pharmaceuticals (Basel). 2025;18(7):1024. https://doi.org/10.3390/ph18071024Cova T, Vitorino C, Ferreira M, et al. Artificial intelligence and quantum computing as the next pharma disruptors. Methods Mol Biol. 2022;2390:321–47. https://doi.org/10.1007/978-1-0716-1787-8_14Cova T, Vitorino C, Ferreira M, et al. Artificial intelligence and quantum computing as the next pharma disruptors. Methods Mol Biol. 2022;2390:321–47. https://doi.org/10.1007/978-1-0716-1787-8_14Patil RS, Kulkarni SB, Gaikwad VL. Artificial intelligence in pharmaceutical regulatory affairs. Drug Discov Today. 2023;28(9):103700. https://doi.org/10.1016/j.drudis.2023.103700Patil RS, Kulkarni SB, Gaikwad VL. Artificial intelligence in pharmaceutical regulatory affairs. Drug Discov Today. 2023;28(9):103700. https://doi.org/10.1016/j.drudis.2023.103700Ajmal CS, Yerram S, Abishek V, et al. Innovative approaches in regulatory affairs: leveraging artificial intelligence and machine learning for efficient compliance and decisionmaking. AAPS J. 2025;27(1):22. https://doi.org/10.1208/s12248-024-01006-5Ajmal CS, Yerram S, Abishek V, et al. Innovative approaches in regulatory affairs: leveraging artificial intelligence and machine learning for efficient compliance and decisionmaking. AAPS J. 2025;27(1):22. https://doi.org/10.1208/s12248-024-01006-5Macdonald JC, Isom DC, Evans DD, Page KJ. Digital innovation in medicinal product regulatory submission, review, and approvals to create a dynamic regulatory ecosystem — are we ready for a revolution? Front Med (Lausanne). 2021;8:660808. https://doi.org/10.3389/fmed.2021.660808Macdonald JC, Isom DC, Evans DD, Page KJ. Digital innovation in medicinal product regulatory submission, review, and approvals to create a dynamic regulatory ecosystem — are we ready for a revolution? Front Med (Lausanne). 2021;8:660808. https://doi.org/10.3389/fmed.2021.660808Nene L, Flepisi BT, Brand SJ, et al. Evolution of drug development and regulatory affairs: The demonstrated power of artificial intelligence. Clin Ther. 2024;46(8):e6–14. https://doi.org/10.1016/j.clinthera.2024.05.012Nene L, Flepisi BT, Brand SJ, et al. Evolution of drug development and regulatory affairs: The demonstrated power of artificial intelligence. Clin Ther. 2024;46(8):e6–14. https://doi.org/10.1016/j.clinthera.2024.05.012Sharma K, Manchikanti P. Regulation of artificial intelligence in drug discovery and health care. Biotechnol Law Rep. 2020;39(5):371–80. https://doi.org/10.1089/blr.2020.29183.ksSharma K, Manchikanti P. Regulation of artificial intelligence in drug discovery and health care. Biotechnol Law Rep. 2020;39(5):371–80. https://doi.org/10.1089/blr.2020.29183.ksFu L, Jia G, Liu Z, et al. The applications and advances of artificial intelligence in drug regulation: a global perspective. Acta Pharmaceutica Sinica B. 2025;15(1):1–14. https://doi.org/10.1016/j.apsb.2024.11.006Fu L, Jia G, Liu Z, et al. The applications and advances of artificial intelligence in drug regulation: a global perspective. Acta Pharmaceutica Sinica B. 2025;15(1):1–14. https://doi.org/10.1016/j.apsb.2024.11.006Gude S, Gude YS. The synergy of artificial intelligence and machine learning in revolutionizing pharmaceutical regulatory affairs. Transl Regul Sci. 2024;6(2):37–45. https://doi.org/10.33611/trs.2024-005Gude S, Gude YS. The synergy of artificial intelligence and machine learning in revolutionizing pharmaceutical regulatory affairs. Transl Regul Sci. 2024;6(2):37–45. https://doi.org/10.33611/trs.2024-005Higgins DC, Johner C. Validation of artificial intelligence containing products across the regulated healthcare industries. Ther Innov Regul Sci. 2023;57(4):797–809. https://doi.org/10.1007/s43441-023-00530-4Higgins DC, Johner C. Validation of artificial intelligence containing products across the regulated healthcare industries. Ther Innov Regul Sci. 2023;57(4):797–809. https://doi.org/10.1007/s43441-023-00530-4Kandhare P, Kurlekar M, Deshpande T, Pawar A. Artificial intelligence in pharmaceutical sciences: a comprehensive review. Med Nov Technol Devices. 2025;27:100375. https://doi.org/10.1016/j.medntd.2025.100375Kandhare P, Kurlekar M, Deshpande T, Pawar A. Artificial intelligence in pharmaceutical sciences: a comprehensive review. Med Nov Technol Devices. 2025;27:100375. https://doi.org/10.1016/j.medntd.2025.100375Oualikene-Gonin W, Jaulent MC, Thierry JP, et al. Artificial intelligence integration in the drug lifecycle and in regulatory science: policy implications, challenges and opportunities. Front Pharmacol. 2024;15:1437167. https://doi.org/10.3389/fphar.2024.1437167Oualikene-Gonin W, Jaulent MC, Thierry JP, et al. Artificial intelligence integration in the drug lifecycle and in regulatory science: policy implications, challenges and opportunities. Front Pharmacol. 2024;15:1437167. https://doi.org/10.3389/fphar.2024.1437167Viswanath S, Fennell JW, Balar K, Krishna P. An industrial approach to using artificial intelligence and natural language processing for accelerated document preparation in drug development. J Pharm Innov. 2021;16(2):302–16. https://doi.org/10.1007/s12247-020-09449-xViswanath S, Fennell JW, Balar K, Krishna P. An industrial approach to using artificial intelligence and natural language processing for accelerated document preparation in drug development. J Pharm Innov. 2021;16(2):302–16. https://doi.org/10.1007/s12247-020-09449-xSaraswat N, Li C, Jiang M. Identifying the question similarity of regulatory documents in the pharmaceutical industry by using the recognizing question entailment system: evaluation study. JMIR AI. 2023;2:e43483. https://doi.org/10.2196/43483Saraswat N, Li C, Jiang M. Identifying the question similarity of regulatory documents in the pharmaceutical industry by using the recognizing question entailment system: evaluation study. JMIR AI. 2023;2:e43483. https://doi.org/10.2196/43483Sapkota K, Aldea A, Younas M, et al. Automating the semantic mapping between regulatory guidelines and organizational processes. Serv Oriented Comput Appl. 2016;10(4):365–89. https://doi.org/10.1007/s11761-016-0197-2Sapkota K, Aldea A, Younas M, et al. Automating the semantic mapping between regulatory guidelines and organizational processes. Serv Oriented Comput Appl. 2016;10(4):365–89. https://doi.org/10.1007/s11761-016-0197-2Anderson C, Algorri M, Abernathy MJ. Real-time algorithmic exchange and processing of pharmaceutical quality data and information. Int J Pharm. 2023;645:123342. https://doi.org/10.1016/j.ijpharm.2023.123342Anderson C, Algorri M, Abernathy MJ. Real-time algorithmic exchange and processing of pharmaceutical quality data and information. Int J Pharm. 2023;645:123342. https://doi.org/10.1016/j.ijpharm.2023.123342Dalal A, Ranjan S, Bopaiah Y, et al. Text summarization for pharmaceutical sciences using hierarchical clustering with a weighted evaluation methodology. Sci Rep. 2024;14(1):20149. https://doi.org/10.1038/s41598-024-70618-wDalal A, Ranjan S, Bopaiah Y, et al. Text summarization for pharmaceutical sciences using hierarchical clustering with a weighted evaluation methodology. Sci Rep. 2024;14(1):20149. https://doi.org/10.1038/s41598-024-70618-wLegehar A, Xhaard H, Ghemtio L. IDAAPM: Integrated database of ADMET and adverse effects of predictive modeling based on FDA approved drug data. J Cheminformatics. 2016;8:33. https://doi.org/10.1186/s13321-016-0141-7Legehar A, Xhaard H, Ghemtio L. IDAAPM: Integrated database of ADMET and adverse effects of predictive modeling based on FDA approved drug data. J Cheminformatics. 2016;8:33. https://doi.org/10.1186/s13321-016-0141-7Whitehead TM, Strickland J, Conduit GJ, et al. Quantifying the benefits of imputation over QSAR methods in toxicology data modeling. J Chem Inf Model. 2024;64(7):2624–36. https://doi.org/10.1021/acs.jcim.3c01695Whitehead TM, Strickland J, Conduit GJ, et al. Quantifying the benefits of imputation over QSAR methods in toxicology data modeling. J Chem Inf Model. 2024;64(7):2624–36. https://doi.org/10.1021/acs.jcim.3c01695Siah KW, Kelley NW, Ballerstedt S, et al. Predicting drug approvals: the Novartis data science and artificial intelligence challenge. Patterns (NY). 2021;2(8):100312. https://doi.org/10.1016/j.patter.2021.100312Siah KW, Kelley NW, Ballerstedt S, et al. Predicting drug approvals: the Novartis data science and artificial intelligence challenge. Patterns (NY). 2021;2(8):100312. https://doi.org/10.1016/j.patter.2021.100312Boonsom S, Chamnansil P, Boonseng S, Srisongkram T. ToxSTK: A multi-target toxicity assessment utilizing molecular structure and stacking ensemble learning. Comput Biol Med. 2025;185:109480. https://doi.org/10.1016/j.compbiomed.2024.109480Boonsom S, Chamnansil P, Boonseng S, Srisongkram T. ToxSTK: A multi-target toxicity assessment utilizing molecular structure and stacking ensemble learning. Comput Biol Med. 2025;185:109480. https://doi.org/10.1016/j.compbiomed.2024.109480Lysenko A, Sharma A, Boroevich KA, Tsunoda T. An integrative machine learning approach for prediction of toxicity-related drug safety. Life Sci Alliance. 2018;1(6):e201800098. https://doi.org/10.26508/lsa.201800098Lysenko A, Sharma A, Boroevich KA, Tsunoda T. An integrative machine learning approach for prediction of toxicity-related drug safety. Life Sci Alliance. 2018;1(6):e201800098. https://doi.org/10.26508/lsa.201800098Gray M, Samala R, Liu Q, et al. Measurement and mitigation of bias in artificial intelligence: a narrative literature review for regulatory science. Clin Pharmacol Ther. 2024;115(4):687–97. https://doi.org/10.1002/cpt.3117Gray M, Samala R, Liu Q, et al. Measurement and mitigation of bias in artificial intelligence: a narrative literature review for regulatory science. Clin Pharmacol Ther. 2024;115(4):687–97. https://doi.org/10.1002/cpt.3117Ribeiro LC, Muzaka V. Needle in a haystack: Harnessing AI in drug patent searches and prediction. PloS One. 2024;19(12):e0311238. https://doi.org/10.1371/journal.pone.0311238Ribeiro LC, Muzaka V. Needle in a haystack: Harnessing AI in drug patent searches and prediction. PloS One. 2024;19(12):e0311238. https://doi.org/10.1371/journal.pone.0311238Williams J, Boyce D, Collu G, et al. Generative AI: A generation-defining shift for biopharma regulatory affairs. Nat Rev Drug Discov. 2025;24(9):651–2. https://doi.org/10.1038/d41573-025-00089-9Williams J, Boyce D, Collu G, et al. Generative AI: A generation-defining shift for biopharma regulatory affairs. Nat Rev Drug Discov. 2025;24(9):651–2. https://doi.org/10.1038/d41573-025-00089-9Sharma S, Mathure D, Dingankar S, Dhapte-Pawar V. In pursuit of software solutions for pharmaceutical regulatory affairs: Insights and trends. Ann Pharm Fr. 2025;83(6):1053–61. https://doi.org/10.1016/j.pharma.2025.05.005Sharma S, Mathure D, Dingankar S, Dhapte-Pawar V. In pursuit of software solutions for pharmaceutical regulatory affairs: Insights and trends. Ann Pharm Fr. 2025;83(6):1053–61. https://doi.org/10.1016/j.pharma.2025.05.005Niazi SK. Regulatory perspectives for AI/ML implementation in pharmaceutical GMP environments. Pharmaceuticals (Basel). 2025;18(6):901. https://doi.org/10.3390/ph18060901Niazi SK. Regulatory perspectives for AI/ML implementation in pharmaceutical GMP environments. Pharmaceuticals (Basel). 2025;18(6):901. https://doi.org/10.3390/ph18060901Syed FM, Faiza Kousar ES. AI in securing pharma manufacturing systems under GxP compliance. International Journal of Machine Learning Research in Cybersecurity and Artificial Intelligence. 2024;15(1):448–72.Syed FM, Faiza Kousar ES. AI in securing pharma manufacturing systems under GxP compliance. International Journal of Machine Learning Research in Cybersecurity and Artificial Intelligence. 2024;15(1):448–72.

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