AutoEncoder-Based Computational Framework for Tumor Microenvironment Decomposition and Biomarker Identification in...

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Melanoma prognosis depends on tumor-intrinsic and extrinsic factors.
• Understanding these factors is crucial for effective treatment strategies.

Purpose of the Study:

• To identify robust gene signatures capturing tumor-intrinsic and extrinsic features in melanoma.
• To evaluate the prognostic and predictive value of these signatures.

Main Methods:

• Systematic analysis of The Cancer Genome Atlas (TCGA) melanoma RNA-seq data.
• Application of an AutoEncoder-based method to identify key gene expression nodes.
• Development of tumor-intrinsic (TI) and tumor-extrinsic (TE) signatures.

Main Results:

• Identified two prognostic gene signatures: TI (MYC pathway activity) and TE (cytotoxic immune cell pathways).
• Validated signatures in independent melanoma datasets, showing prognostic capability.
• Demonstrated TE signature association with positive response to immunotherapies.

Conclusions:

• Developed a novel computational framework for analyzing tumor-intrinsic and -extrinsic features.
• Identified TI and TE signatures as robust prognostic and predictive biomarkers for melanoma.
• These signatures can guide treatment decisions and predict immunotherapy response.