Related Experiment Video
Updated: Sep 11, 2025
10:33
Molecular Profiling of the Invasive Tumor Microenvironment in a 3-Dimensional Model of Colorectal Cancer Cells and Ex vivo Fibroblasts
Published on: April 29, 2014
11.2K
Histology-Based Virtual RNA Inference Identifies Pathways Associated With Metastasis Risk in Colorectal Cancer
Gokul Srinivasan1, Minh-Khang Le1, Zarif Azher2
1Departments of Pathology and Laboratory Medicine and Computational Biomedicine, Cedars-Sinai Medical Center, Los Angeles, California.
Summary
Virtual RNA Inference (VRI) uses standard H&E images to predict spatial transcriptomics data for colorectal cancer research. This method offers a cost-effective way to analyze the tumor microenvironment and identify prognostic markers.
Area of Science:
- Oncology
- Computational Biology
- Pathology
Background:
- Colorectal cancer (CRC) poses a significant health burden, necessitating advancements in screening and treatment.
- The tumor microenvironment (TME) is crucial for CRC progression and metastasis but is difficult to analyze spatially.
- Current spatial transcriptomics (ST) technologies are expensive and have limitations for large-scale studies.
Purpose of the Study:
- To refine and implement Virtual RNA Inference (VRI) for deriving ST-level molecular information from H&E-stained images.
- To enable high-resolution TME analysis without the need for costly ST profiling.
- To identify prognostic biomarkers for CRC metastasis and recurrence.
Main Methods:
- Trained VRI models on the largest matched CRC ST dataset (45 patients, >300,000 spots).
- Utilized state-of-the-art deep learning architectures (UNI, ResNet-50, ViT, VMamba).
- Validated VRI-derived gene signatures against direct ST and assessed cell-type proportion estimation.
Main Results:
- Achieved a median Spearman's correlation of 0.546 between predicted and measured spot-level gene expression.
- VRI-derived signatures showed strong concordance with ST-generated signatures for tissue regions.
- Identified VRI-derived gene signatures significantly associated with CRC metastasis status in a large cohort.
Conclusions:
- VRI can infer a wide range of "histology-associated" biological pathways at near-cellular resolution from H&E images.
- This approach bypasses the need for expensive ST profiling, accelerating translational CRC research.
- VRI has the potential to expand TME phenotyping from standard pathology slides for large-scale studies.
