The researchers have developed a novel deep learning framework that integrates high-resolution pathology images with spatial transcriptomics and proteomics to reveal the complex intra-tumor heterogeneity. The strategy, called Deep Visual Spatial Transcriptomics and Proteomics (DVSTP), enables three-dimensional reconstruction of entire tumors and identifies distinct molecular subtypes that correlate with immune cell infiltration patterns. These findings, published in Science Bulletin, could transform how researchers understand tumor evolution and guide precision cancer treatments.
Colorectal cancer remains one of the most challenging malignancies to treat, partly because tumors are not uniform masses but complex ecosystems composed of diverse cell populations with distinct behaviors. This intra-tumor heterogeneity—the variation in gene expression and protein levels across different regions of the same tumor—critically determines how cancers progress and respond to therapy. However, conventional sequencing methods that homogenize tumor tissue lose the spatial context essential for understanding these cellular interactions.
“Bulk sequencing gives us an average picture, like blending an entire landscape into a single color,” explains the research team from Union Hospital, Tongji Medical College, Huazhong University of Science and Technology. “We needed a way to see both the molecular details and where they occur in the tissue.”
To address this challenge, the researchers developed DVSTP, a strategy that combines three data layers: high-resolution histopathology images, spatial transcriptomics, and mass spectrometry-based spatial proteomics. Using tissue samples from 123 colorectal cancer patients, the team first trained a deep learning model to identify cell types from H&E stained images, achieving 94% accuracy in distinguishing malignant cells, immune cells, and other stromal components.
The researchers then applied this approach to serial sections from two distinct sites within a single stage Ⅱ colorectal tumor, analyzing 380 slices and reconstructing the three-dimensional tumor architecture. “This allowed us to see how different cell populations are organized in space and how they interact across the entire tumor volume,” says the lead author.
Comparing spatial transcriptomics and proteomics data: the correlation between mRNA and protein levels was surprisingly modest, with an average Spearman correlation of only 0.37. This supports the growing recognition that gene expression is tightly regulated and that protein levels cannot be reliably inferred from transcript data alone.
“The mRNA-protein relationship is not a simple linear translation,” the researchers note. “This is why integrating direct protein measurements is so crucial for understanding functional cell states.”
The spatial proteomics analysis identified 2,805 proteins and revealed significant heterogeneity in protein expression across tumor regions. Four distinct tumor subtypes emerged from the protein data, each with unique molecular signatures.
One of the most promising aspects of the study was the demonstration that computational analysis of standard pathology images alone could predict molecular profiles. The team’s deep learning model achieved an area under the curve (AUC) of 0.718 for predicting protein expression from H&E images, improving to 0.755 when combining images with transcriptomics data.
“This suggests that the morphological patterns pathologists have used for decades contain rich molecular information that we can now decode with artificial intelligence,” the researchers explain. The approach was validated using data from The Cancer Genome Atlas, confirming its generalizability across independent patient cohorts.
Among the heterogeneous protein patterns identified, Serine/Arginine-Rich Splicing Factor 6 (SRSF6) stood out due to its pronounced spatial variation and strong association with immune exclusion—regions of the tumor with reduced immune cell infiltration. SRSF6-high areas showed significantly lower presence of CD4 + and CD8 + T cells, suggesting this protein creates an immunosuppressive microenvironment.
The researchers validated these findings through experiments in cell lines and mouse models. Srsf6 overexpression promoted cancer cell migration and tumor growth while reducing T cell infiltration, whereas knockdown had the opposite effects. In human patient samples, high SRSF6 expression correlated with poorer overall survival.
“SRSF6 appears to be a master regulator that shapes both tumor cell behavior and the immune landscape,” the team concludes. “Its spatial heterogeneity within tumors may explain why some regions evade immune detection.”
The researchers emphasize that current spatial transcriptomics and proteomics technologies remain limited by resolution and cost. While emerging platforms offer improved performance, they are not yet widely accessible. The DVSTP strategy addresses this by using computational deconvolution to enhance effective resolution and integrating readily available pathology images with spatial omics data.
The approach also enables three-dimensional reconstruction of entire tumors, revealing spatial infiltration patterns that would be invisible in two-dimensional sections. This capability could help clinicians identify aggressive tumor regions and predict therapeutic responses based on the spatial organization of immune cells.
Science Bulletin
News article