The massive amount of health-related data such as omics, imaging, drug, clinical and diverse ontology data enables the unraveling of the underlining mechanisms of diseases and the finding of optimal treatments. However, data alone cannot provide better understanding. Data must be analyzed systematically and without bias in order to bring about hypotheses for further clinical validations. In this presentation, we focus on data analytics for temporal data. In particular, we model tumor progression in non-small cell lung cancer (NSCLC) via graph comparisons among time-specificgraphs. Results showed that the identified temporal structures capture molecular mechanisms in tumor progression in NSCLC. In particular, we identified major histocompatibility complex of class II and proteasome temporal structures that capture mechanisms related to carcinogenesis and tumor progression respectively. Importantly, while we modeled tumor progression in NSCLC, the method is generic, and can be applied to other disease temporal datasets.