Tumors are known for profound heterogeneity, especially in their genome. To further develop personalized medicine, there have been many studies elucidating the complexity of the cancer genome. In this seminar, I would like to present the computational efforts in characterizing tumor heterogeneity, encompassing 1) inter-tumor heterogeneity; 2) temporal heterogeneity, and 3) intra-tumoral heterogeneity. To characterize inter-tumor heterogeneity, we performed multi-omics approaches to identify subtypes of hormone-dependent tumors and developed a robust machine-learning model to predict drug responses. To address temporal heterogeneity, we infer a clonal association between primary and recurrent tumors based on the copy number profiles. The clonal association can serve as an indication of the treatment applied to the primary disease. Finally, to address intra-tumoral heterogeneity, we developed a novel statistical method that deconvolutes standard RNA-seq data. The advanced techniques allow users to address the tumor microenvironment without single-cell RNA-seq data. Furthermore, we are working on mathematical modeling of cellular distribution in tumor microenvironment to quantitatively characterize cellular interplay. Based on all of these efforts, we are trying to find the clinical relevance to further optimize patient care.