Despite advances in cancer molecular profiling, successful therapeutic development has been hindered by challenges in identifying tumor-specific mechanisms that can be targeted without consequence to healthy tissue. Discrimination between tumor and host cells that comprise the tumor microenvironment remains a difficult yet important task for defining tumor cell signatures. Correspondingly, a computational framework capable of accurately distinguishing tumor from non-tumor cells has yet to be developed. Cell annotation algorithms are largely unable to assign integrated genomic and transcriptional profiles to single cells on a cell-by-cell basis. To address this, we developed the Single Cell Rule Association Mining (SCRAM) tool that integrates RNA-inferred genomic alterations with co-occurring cell type transcriptional signatures for individual cells. Applying our pipeline to human and mouse glioma, we identified tumor cell trajectories that recapitulate temporally-restricted developmental paradigms and feature unique co-occurring genomic and transcriptomic identities. Specifically, we describe and validate two previously unreported tumor cell populations with immune and neuronal signatures as hallmarks of human glioma subtypes. In vivo modeling revealed an immune-like tumor cell population can direct CD8+ T cell responses and survival outcomes. In parallel, electrophysiology and Patch-seq studies in human tumors confirmed a frequent subset of neuronal-like glioma cells that fire action potentials but retain the morphology of glia. These collective studies report the existence of new glioma cell types with functional properties akin to their non-tumor analogs and demonstrate the ability of SCRAM to identify and characterize these cell types in unprecedented detail.