The majority of advanced melanoma patients who are treated with standard of care anti-PD-1 based immunotherapies will have a disease recurrence and require an alternative therapy to effectively treat their disease. The Personalised Immunotherapy Program (PIP), has developed machine learning based predictive models to identify patients who will not respond to either anti-PD-1 monotherapy of combinations with anti-CTLA-4. The models have been developed and validation in over 500 patients with stage III/IV metastatic melanoma who received anti-PD-1 based immunotherapies. Biomarkers including somatic mutation (Qiaseq TMB IO, Qiagen), gene expression (Pancancer 360 IO, Nanostring) and tumour immune profiling (Opal mIHC, Akoya Bioscience) has been performed to generate these prediction. This presentation will cover the development and validation of these muti-omic models and outline the progress towards clinical implementation of the approach within a prospective cohort study (PIP-PREDICT), which is running in oncology clinics in real-time.