Gene Swin transformer: new deep learning method for colorectal cancer prognosis using transcriptomic data

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Transcriptome sequencing is crucial for tumor research but generates complex data.
• Identifying clinically relevant information from gene expression is challenging.

Purpose of the Study:

• To develop a novel method, Gene Swin Transformer, for analyzing transcriptomic data.
• To predict colorectal cancer prognosis using deep learning on transformed transcriptomic data.

Main Methods:

• Transcriptomic data converted into Synthetic Image Elements (SIEs).
• Deep learning models (BeiT, ResNet, Swin Transformer, ViT Transformer) trained and evaluated.
• Performance compared using precision, recall, F1 scores, and AUC values.

Main Results:

• The enhanced Swin-T model demonstrated superior performance in prognostic prediction.
• Achieved weighted precision (0.708), recall (0.692), and F1 (0.705).
• Identified PEX10 gene as a significant prognostic marker via attention analysis and bioinformatics.

Conclusions:

• Gene Swin model effectively transforms RNA sequencing data into SIEs for prognosis prediction.
• The approach supports a data-driven, automated model for RNA sequencing analysis.
• Presents a novel clinical strategy for cancer prognosis forecasting.