Detecting phenotype-specific tumor microenvironment by merging bulk and single cell expression data to spatial transcriptomics

In addressing the limitations of current multimodal analysis methods that largely ignore phenotypic data, leading to a lack of biological interpretability at the phenotypic level, we developed the Single-Cell and Tissue Phenotype prediction (SCTP), a deep-learning-based multimodal fusion framework. SCTP can simultaneously detect phenotype-specific cells and characterize the tumor microenvironment of pathological tissue by integrating essential information from the bulk sample phenotype, the composition of individual cells, and the spatial distribution of cells. Upon evaluating SCTPs efficiency and robustness against traditional analytical methods, we developed a specialized model, SCTP-CRC, tailored for colorectal cancer (CRC). This model integrates RNA-seq, scRNA-seq, and spatial transcriptomic data to offer a better understanding of CRC. SCTP-CRC has proven effective in accurately identifying tumor-associated cells and clusters and continuously defines boundary regions as well as the spatial organization of the entire tumor microenvironment. This enables a detailed depiction of cellular communication networks, capturing the dynamic shifts that occur during tumor progression. Furthermore, SCTP-CRC extends to the identification of abnormal sub-regions in the early state of CRC and uncovers potential early-warning signature genes such as MMP2, IGKC, and PIGR. These biomarkers are not only important in recognizing the onset of CRC but may also play a crucial role in differentiating between CRC-derived liver metastases and primary liver tumors. SCTP stands as a transformative framework, offering a deeper understanding of the tumor microenvironment through its ability to quantitatively characterize cancers fundamental traits and dissect the intricate molecular and cellular interactions at play. This comprehensive insight supports the early diagnosis and enables personalized treatment strategies, marking a significant stride toward improving patient outcomes and tailoring therapies to individual disease profiles.