MTMol-GPT is a transformer-based generative adversarial imitation learning model for de novo multi-target molecular generation. It uses a dual discriminator model with Inverse Reinforcement Learning (IRL) to generate molecules targeting multiple proteins simultaneously. The model pre-trains on the ChEMBL database to ensure generated molecules are valid and drug-like. It employs a dual contrastive discriminator to estimate the realness of generated molecules and uses a replay buffer to store molecules from different targets. MTMol-GPT outperforms existing methods in generating valid, novel, and effective multi-target molecules for complex diseases. Molecular docking and pharmacophore mapping experiments show the generated molecules have good drug-like properties and potential for neuropsychiatric interventions. The model's generalizability is demonstrated through a case study on breast cancer drug design. MTMol-GPT provides a rapid and accurate method for generating high-quality multi-target molecules, enhancing potential complex disease therapeutics. The model uses SMILES and SELFIES representations for molecular generation and shows superior performance in various metrics. It also demonstrates the ability to generate molecules targeting multiple proteins simultaneously, with high similarity to training data and good binding affinity to target proteins. The model's results indicate its potential for drug discovery and development.MTMol-GPT is a transformer-based generative adversarial imitation learning model for de novo multi-target molecular generation. It uses a dual discriminator model with Inverse Reinforcement Learning (IRL) to generate molecules targeting multiple proteins simultaneously. The model pre-trains on the ChEMBL database to ensure generated molecules are valid and drug-like. It employs a dual contrastive discriminator to estimate the realness of generated molecules and uses a replay buffer to store molecules from different targets. MTMol-GPT outperforms existing methods in generating valid, novel, and effective multi-target molecules for complex diseases. Molecular docking and pharmacophore mapping experiments show the generated molecules have good drug-like properties and potential for neuropsychiatric interventions. The model's generalizability is demonstrated through a case study on breast cancer drug design. MTMol-GPT provides a rapid and accurate method for generating high-quality multi-target molecules, enhancing potential complex disease therapeutics. The model uses SMILES and SELFIES representations for molecular generation and shows superior performance in various metrics. It also demonstrates the ability to generate molecules targeting multiple proteins simultaneously, with high similarity to training data and good binding affinity to target proteins. The model's results indicate its potential for drug discovery and development.