Genomic studies have revealed the heterogeneity of breast cancer and identified "intrinsic molecular subtypes" with significant difference in incidence, survival and therapeutic response. Investigation of their clinical implications is critical for personalized therapeutics and drug development. The characteristics of cancer genomics require special considerations in the application of laboratory and computational approaches. Therefore, my research explored the use of two technologies, Genetically Engineered Mouse Model (GEMM) and RNA-sequencing (RNA-seq), to facilitate the translation of cancer biology into clinical knowledge. One powerful GEMM, the p53-null transplant model, was identified as a heterogeneous model that gave rise to multiple subtypes, including Basal-like, Luminal and Claudin-low. Molecular characterization identified genetic signatures of GEMM and its human counterpart and distinct genomic DNA copy number changes associated with each subtype. The analysis on the Claudin-low p53-null tumors showed that they have high expression of epithelial-to-mesenchymal transition genes and are enriched for tumor initiating cells, therefore revealing the stem-cell characteristics of Claudin-low. The utility of GEMM also involves preclinical drug efficacy testing. We evaluated the efficacy of four chemotherapeutic and/or targeted anti-cancer drugs using three well-established mouse models that recapitulate three human subtypes: Basal-like, Luminal B and Claudin-low. Additionally, we identified two gene signatures that predicted pathological complete response to neoadjuvant anthracycline/taxane therapy in humans. The predictive significance was further validated in two large, independent cohorts of human patients, suggesting the possibility of deriving new biomarkers for humans from analysis of GEMM genomic data. Another resource of cancer genomics is the formalin-fixed paraffin-embedded (FFPE) samples. Though RNA-seq has been adopted by many studies, the mRNA enrichment protocol (mRNA-Seq) to remove rRNA restricted its utility in FFPE. We examined two rRNA depletion protocols on paired fresh-frozen (FF) and FFPE samples, and compared them with mRNA-seq and DNA microarray. We demonstrated that Ribo-Zero-Seq provides equivalent rRNA removal efficiency and coverage uniformity. Both protocols have consistent transcript quantification using FF and FFPE, suggesting that RNA-seq can be performed on FFPE. My work uses multiple genomic data types to identify murine models and to develop new protocols for the development and evaluation of new biomarkers for human breast cancer patients.