T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive malignancy with diverse oncogenic drivers, the relative contributions of which remain obscure. Enforced expression of select oncogenes is sufficient to generate T-ALL in mice, however, mouse models are limited in their value for understanding human disease. We recently developed the first “synthetic” model of human T-ALL by lentiviral transduction of normal human CD34+ cord blood progenitors with a combination of four known T-ALL oncogenes: NOTCH1, LMO2, TAL1, and BMI1 (NLTB). These cells expand robustly in culture and produce aggressive, serially transplantable T-ALL in immunodeficient mice. Additionally, we found that LTB alone fails to perform in vitro or in vivo, highlighting an essential role of NOTCH1. To determine the minimal complement of oncogenes required to generate de novo human T-ALL, we executed a “leave-one-out” and “leave-two-out” strategy. We scored the following oncogene combinations in vitro and in vivo: NLB, NLT, NTB, NL, NT, and NB. We found that the various oncogene combinations yielded a spectrum of aberrant phenotypes affecting cell differentiation and proliferation, and a subset produced aggressive leukemias in mice. We also performed RNA-seq on non-leukemogenic and preleukemic cell populations, and fully transformed leukemic cells to define gene expression programs necessary for cellular transformation. Our approach allows us to attribute specific gene programs and cellular phenotypes to each oncogene individually and in combination. Our synthetic approach is flexible, reproducible, and experimentally tractable, allowing functional testing of individual genetic variants and can also serve as a customizable platform for testing of targeted therapeutics.