Abstract

Tumor growth and resistance arise from the interplay between cell-cycle dysregulation and the spatial organization of the tumor microenvironment (TME). While spatial transcriptomics now enables molecular profiling of intact tissues, it captures only static molecular states, making it challenging to reconstruct dynamic processes such as proliferation. Here, we develop SpaceCycle, a computational framework that infers the continuous cell-cycle phase and associated oscillatory gene expression dynamics from spatial transcriptomic data. Using human melanoma, breast, and lung tumors, we map both discrete and continuous cell-cycle states across entire tissue sections to reveal how proliferative activity is spatially structured within the TME. We find that cycling and non-cycling cells form distinct spatial niches reflecting differences in vascularization, immune composition, and cellular density. Continuous-phase inference further exposes tumor-specific variations in cell cycle phase durations and uncovers oscillatory programs. Together, our results provide a spatiotemporal view of cell-cycle organization in human tumors and establish a general framework for detecting dynamic transcriptional programs and spatial proliferation patterns from static tissue data.

Manuscript