Linking spatial and temporal profiling to clinical settings through advanced machine learning

Leveraging advancements in spatially resolved single-cell profiling, our study enhances understanding of glioma evolution and the complex tumor microenvironment. Through spatially resolved multi-omic analysis of glioma samples from different treatments and stages, we've mapped glioblastoma's architectural changes over time, highlighting mechanisms behind therapy resistance. Our use of array-based and in-situ transcriptomics across treatment types revealed key transcriptional programs related to resistance, showing significant inter-patient heterogeneity and unique resistance pathways in recurrent tumors. We identified  adaptations in the neural and immune microenvironment shifts post-immunotherapy, indicating mechanisms of resistance development. By integrating spatial transcriptomics and clinical data into graph-neural networks, we've created models predicting treatment responses, offering new insights for precision oncology.