Breast cancer is a highly heterogeneous disease, exhibiting both inter- and intra- tumour heterogeneity at genomic and phenotypic levels. This remains a key limitation in the treatment of the disease. While recent advances have dramatically improved our understanding and treatment strategies for specific subgroups of patients, those with triple negative breast cancer remain limited to few therapeutic options, due to a lack of actionable targets. Developments in our understanding of the DNA damage response in breast cancer have unveiled an unprecedented opportunity to treat this subgroup of patients, who commonly harbour alterations in related pathways. Accelerated by the discovery of breast cancer predisposition genes BRCA1 and BRCA2, a number of compounds which target the DNA damage response are entering preclinical drug development. However, the success of new therapies is dependent upon preclinical models which mirror clinical drug responses and accurately capture the heterogeneity of this complex disease. Patient-derived tumour xenografts (PDTXs) have been found to recapitulate the main genomic and histological features of breast cancer and reflect the diversity of drug responses observed in the clinic. This thesis outlines the key findings on the development of an avatar co-clinical trial platform to assess the concordance with clinical drug responses on a matched patient/PDTX basis. However, as with many approved therapies, an understanding of the dynamics of drug response and the evolutionary principles underpinning the disease remain at the forefront of biological interest and are essential to prevent, or anticipate, the development of drug resistance. In order to approach this question, we leveraged the preclinical platform and used single cell technologies to shed light on the dynamics of drug response and mechanisms of drug tolerance, to drugs targeting the DNA damage response in breast cancer.