The discovery of new chemical compounds with specified properties is a challenging problem in drug development. Many studies encode molecules as string representations derived from molecular graphs; however, this approach is computationally expensive and does not readily extend to general (non-molecular) graphs. Advances in graph deep learning make it possible to train generative models directly on graph representations, avoiding costly search in the discrete and extremely large space of chemical structures. MolGAN is a family of implicit models for generating small molecular graphs that combines generative adversarial networks (GANs) with reinforcement learning (RL) to produce molecules with target chemical properties. This review considers four variants: the baseline MolGAN (up to 9 atoms), Large MolGAN (up to 20 atoms) with a graph-expansion mechanism that reduces the generation of disconnected graphs, a MolGAN variant based on WGAN as a more stable alternative to standard GAN training, and a hybrid MolGAN incorporating quantum computing modules. The paper describes the model architectures, compares their performance on established benchmarks, and discusses limitations and directions for future research.
| Mualliflar | Adilova, F.T., Davronov, R.R., Адылова, Ф.Т., Давронов, Р.Р. |
|---|---|
| Jurnal | Ҳисоблаш ва амалий математика муаммолари |
| Nashr sanasi | 2026-03-07 |
| Son | 1 |
| Betlar | 123-142 |
| Til | Rus |
| DOI | 10.71310/pcam.1_71.2026.11 |
DOI: 10.71310/pcam.1_71.2026.11 · Maqolaning asl sahifasi
reinforcement learning, WGAN, implicit generative models, graph connectivity enforcement, quantum–classical hybrid models, molecular design, molecule generation benchmarks, обучение с подкреплением, WGAN, неявные генеративные модели, обеспечение связности графов, квантово-классические гибридные модели, молекулярный дизайн, бенчмарки генерации молекул
This paper examines the dynamic modes of the fractional Zeeman oscillator. The fractional Zeeman oscillator is a system of two ordinary differential equations with fractional derivatives, understood in the…
The solutions of boundary value problems associated with two-parameter singularly perturbed differential equations are well known to exhibit the formation of two distinct boundary layers, typically occurring near the…
This paper examines a boundary value problem for a degenerate elliptic equation in a planar domain bounded by a line segment and an analytic curve. The primary objective is to construct an explicit analytical solution…
The method of displaced nodes is applied when solving the Dirichlet problem for the Poisson equation in a rectangular domain. By approximating the Laplace operator using moving nodes, we obtain an approximate analytical…
This research develops a mathematical formulation and numerical solution for the spatial-temporal distribution of groundwater head in heterogeneous porous media. Based on Darcy’s and mass conservation laws, a 3D…
The problem of constructing optimal quadrature formulas for approximate calculation of definite integrals and approximation of functions is one of the important problems of computational mathematics. These problems have…
Recent years have witnessed significant climatic changes and increasing environmental pressure globally, including in Uzbekistan, necessitating an objective regional assessment. This research develops an approach for…
In the early 1960s, seminal results in differential game theory, where the game is described by an ordinary differential equation in a finite-dimensional space, were obtained by academicians L.S. Pontryagin and N.N…
This research presents a multidimensional mathematical model and a robust secondorder numerical algorithm for coupled heat and moisture transfer with pressure-driven gas flow. Utilizing fractional Caputo derivatives (0…
In this article, within the framework of the Saint-Venant compatibility conditions, plane problems of plasticity theory formulated in terms of strains are presented, aimed at investigating the stress-strain state of a…