Prediction of TNFRSF9 expression and molecular pathological features in thyroid cancer using machine learning to...

Area of Science:

• Oncology
• Genomics
• Computational Pathology

Background:

• Tumor necrosis factor receptor superfamily member 9 (TNFRSF9) plays a critical role in thyroid carcinoma (THCA) pathogenesis.
• Predicting TNFRSF9 expression and understanding its molecular mechanisms are crucial for THCA treatment strategies.

Purpose of the Study:

• To develop and validate a Pathomics model for predicting TNFRSF9 expression in THCA.
• To explore the molecular mechanisms and prognostic implications of TNFRSF9 expression in THCA.

Main Methods:

• Utilized The Cancer Genome Atlas (TCGA) data including transcriptome, pathology images, and clinical information.
• Employed image segmentation (OTSU's algorithm) and feature extraction (pyradiomics) for Pathomics analysis.
• Developed a Gradient Boosting Machine (GBM) model with feature selection (mRMR_RFE) to predict TNFRSF9 expression and assess prognostic value.

Main Results:

• High TNFRSF9 expression is associated with poorer progression-free interval (PFI) and serves as an independent risk factor.
• The GBM Pathomics model achieved good prediction efficacy (AUC 0.819, 0.769) and identified nine key pathohistological features.
• Elevated probabilistic pathomics score (PS) correlated with increased risk, enriched pathways, higher TIGIT expression, Tregs infiltration, and more gene mutations.

Conclusions:

• A GBM Pathomics model effectively predicts TNFRSF9 expression levels in THCA using H&E-stained histopathological features.
• The model offers valuable insights into the molecular mechanisms and prognostic significance of TNFRSF9 in thyroid carcinoma.