Анализ решений обратной задачи поиска отношения «структура-активность» (QSAR) на базе глубоких нейронных сетей

Abstract

Artificial Intelligence (AI) is a topic that has been intensively discussed and applied in many areas. Machine Learning (ML) methods are among the approaches of AI, one of which is deep learning (DL) using deep neural networks (Deep Neural Network, DNN). Compared to other machine learning methods, DL shows good efficiency and prospects in the field of image analysis and language processing. In 1989, a group of scientists Baskin I.I., Gordeeva E.V., Devdariani R.O., Zefirov N.S., Palyulin V.A., Stankevich M.I. For the first time, the inverse problem was formulated in the “structure-property” problem, the essence of which is to search for structures of chemical compounds having propertiesthat lie within predetermined intervals. This problem was solved for some examples by reducing it to a combinatorial problem from the field of discrete mathematics by listing graphs whose values of certain invariants lie in given intervals. Much has changed over 30 years, and interest in solving the inverse problem through the use of DL in medical chemistry has again made it relevant. An analysis of the state of DL application for solving the most urgent task of drug development, namely, for the so-called inverse QSAR problem, is the subject of this review.

More by Fatima Adilova

• Graph-based density-variation outlier removal for chronic disease prediction: A leakage-free benchmark and component-wise ablationResults in Engineering · 2026
• Monotonic System Framework for Globally Optimal Structuring of Complex Data2026
• Boosting LLM Adaptation: Superior Stability and Convergence via OLoRA+ Across Diverse NLP BenchmarksInternational Siberian workshops and tutorials on electron devices and materials · 2026
• Генерации графов заданной структуры: от глубоких нейронных сетей к квантовым моделям (на примере создания новых лекарств)Problems of Computational and Applied Mathematics · 2026