MTMol-GPT: De novo multi-target molecular generation with transformer-based generative adversarial imitation learning

• DOI: 10.1371/journal.pcbi.1012229
• Publication Date: 2024-06-26T17:49:37Z
• Authors: Chengwei Ai; Hongpeng Yang; Xiaoyi Liu; Ruihan Dong; Yijie Ding; Fei Guo
• Keywords: Drug target; Discriminator

De novo drug design is crucial in advancing discovery, which aims to generate new drugs with specific pharmacological properties. Recently, deep generative models have achieved inspiring progress generating drug-like compounds. However, the prioritize a single target generation for intervention, neglecting complicated inherent mechanisms of diseases, and influenced by multiple factors. Consequently, developing novel multi-target that simultaneously targets can enhance anti-tumor efficacy address issues related resistance mechanisms. To this issue inspired Generative Pre-trained Transformers (GPT) models, we propose an upgraded GPT model adversarial imitation learning molecular called MTMol-GPT. The generator employs dual discriminator using Inverse Reinforcement Learning (IRL) method concurrently generation. Extensive results show MTMol-GPT generates various valid, novel, effective molecules complex demonstrating robustness generalization capability. In addition, docking pharmacophore mapping experiments demonstrate drug-likeness properties effectiveness generated potentially improve neuropsychiatric interventions. Furthermore, our model’s generalizability exemplified case study focusing on multi-targeted breast cancer. As broadly applicable solution targets, provides insight into future directions potential disease therapeutics high-quality discovery.