Research database

The project 'LangMol - Unlocking Chemical Design with Molecular Language Modelling,' proposed by Dr. Gianvito Grasso, aims to develop a computational platform powered by Natural Language Processing (NLP) for the rational design and development of small molecules and drugs. This platform is designed to interpret and guide users through natural language queries across the complex stages of chemical design. When necessary, the embedded virtual assistant will use a suite of machine learning models, based on molecular and chemical language models, which will be pre-trained and fine-tuned specifically in the LangMol project. By harnessing the capabilities of language models, our aim is to design and develop a computational platform that can evaluate potential drug candidates using multi-task predictions related to ADME properties, protein- binding affinity, toxicity, and drug-likeness. This pioneering approach holds the potential to accelerate the drug discovery process, providing pharmaceutical and biotechnological companies with a robust tool to navigate the complexities of chemical design. Addressing this ambitious goal requires collaborative synergy among key partners: Politecnico di Torino, known for its expertise in machine learning, molecular modeling, and computational biophysics; Scuola Universitaria Professionale della Svizzera Italiana, specializing in machine learning and NLP; and Claranet Italia, a leading European Managed IT Services provider. Building on the revolutionary wave of language model technologies, LangMol innovatively applies these advanced models to molecular simulations, bridging the gap between language processing and drug development. This combination of cutting-edge technology positions the project at the forefront of modern drug discovery, offering a unique and transformative approach to molecular design.