PROJECT SUMMARY – PROJECT 2 We hypothesize that patient-derived models can be used to identify more effective treatments for metastatic breast cancer that can be predicted using molecular biomarkers, and that these models can be used to accurately predict and interrogate treatment resistance. We will use our large collection of PDX, organoid, and tissue slice culture models established from patients at different stages of treatment (including longitudinal samples and primary:metastatic pairs from the same patient) to investigate new therapies. We will capitalize on our new, very extensive dataset of NCI-IND drug responses in tumor organoids isolated from PDX (PDxOs): We have screened nearly 100 PDxO models comprising all breast cancer subtypes, each with 40-50 drugs. Effective “lead” therapies have been validated in vivo in PDX models; there is high concordance between organoid and PDX data. Omics data are available on all of the models, and are being used to identify biomarkers associated with responders or non-responders, for each lead therapy. We will apply new computational algorithms to this rich dataset to prioritize drugs that can be used in combination with standard therapy, or in combination with each other. The primary outcome of this study will be new NCI-IND treatment regimens for metastatic breast cancer, using patient-derived models as a preclinical tool to (1) evaluate the efficacy of many drugs and drug combinations across a large number of different patients’ tumor models, and (2) identify the biomarkers of responding and non- responding breast cancers for new therapies, in order to better inform clinical trial design.