I present two computational studies of gene regulation: the first is a machine learning approach to predict potent shRNAs to knock down specific endogenous mRNAs; the second is an in-depth analysis of the relationship between chromatin accessibility and the functional state of tumor-specific CD8+ T cells. Short-hairpin RNA, or shRNA, are synthetic RNA molecules that can be used to silence mRNA transcripts in a sequence-dependent manner and are used extensively in gain- and loss-of-function genetic studies. In order to predict which shRNA molecules will be potent despite changes in underlying shRNA technology, we developed a cascaded support vector machine, SplashRNA, trained on both types of shRNA backbone, miR-30 and miR-E. This strategy allows us to learn basic shRNA potency rules on our larger but older miR-30 dataset and followed by more precise miR-E-specific rules on our smaller miR-E training set. We demonstrate that SplashRNA outperforms all other shRNA prediction methods while limiting off-target effects through careful curation of mRNA transcript data and we have developed an open-source implementation of this algorithm available at splashrna.mskcc.org. Tumor-specific T cells have been found in patients’ tumors, but these tumors continue to progress, indicating that these T cells are not functional. A subset of patients in this situation have responded to checkpoint blockade therapy, which can rescue silenced T cells, but not all patients are responsive to this treatment and some only respond for a brief period. Here, we investigate the chromatin accessibility landscape of normal and dysfunctional tumor-specific T cells in order to determine the regulatory changes that take place when T cells are in a dysfunction-inducing tumor environment. We computationally identify and pharmacologically validate NFAT and TCF family members as critical in this differentiation to dysfunction. We also identify cell surface markers that differentiate between reprogrammable and fixed cells, a strategy that may be used to determine if patients are candidates for checkpoint blockade therapy.