The goal of our project is building a predictive response algorithm for patients with metastatic lung cancer, exploiting an artificial intelligence platform. It will collect patient information from all areas (clinical, laboratory, radiological, pathological) and analyse them, understanding connections and correlations, both at baseline and at pre-specified timepoints. It would lead to the development of a reliable and constantly evolving predictive score, able to continuously re-weight the importance of each variable as new data come in. Since the greatest clinical need is identifying non-responders to immunotherapy and chemo-immunotherapy combination (30% of all treated patients), these two populations are defined as the starting cohorts (Cohort A, immunotherapy alone, Cohort B, chemo-immunotherapy combinations). For each cohort, three main questions are to be answered: Q1) Early progressors (defined as progressive disease or death within three months of treatment or at first radiological restaging) Q2) Toxicity (with a special focus on severe toxicities G≥3) Q3) Long survivors (defined as patients reaching an overall survival of at least 1.5x median overall survival in registrative trials) The early identification of non-responders, high-risk patients (or on the other hand, long survivors) would help their healthcare planning, providing individualised follow-up strategies or prompting their inclusion in alternative treatments (eg clinical trials). For all cohorts, first data entry will be retrospective and second data entry will be prospective (as validation set).