Gene expression-based consensus molecular subtypes (CMS) provide mechanistic, prognostic, and therapeutic insights in colorectal cancer. However, the original four subtypes were defined using bulk transcriptomic data, such that intra-subtype heterogeneity and cell type contributions are not fully clarified. In their study on page 240, Bhukdee and colleagues employed computational representation learning and single-cell transcriptomic dataset comparisons to impart higher resolution to CMS and their constituent genes and cell types. The authors' analyses produced a 62-gene panel that identifies 3 novel subtypes within 2 existing CMS, and single-cell transcriptomic data imposition onto the panel elucidates tumor and non-tumor cell contributions to the subtypes. The cover image is a 2-dimensional densMAP visualization of the 4 original and 3 novel subtypes using the identified 62-gene panel, where each subtype is designated with a unique color. The presented analyses and gene panel further delineate molecular mechanisms underlying colorectal cancer and may enhance the power of CMS to inform disease prognoses and therapeutic strategies.