Identification of Prognostic Biomarker Candidates Associated With Melanoma Using High-Dimensional Genomic Data

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Metastatic melanoma is a challenging cancer with variable patient survival and chemoresistance.
• Current immunotherapies like checkpoint inhibitors improve response rates but offer limited long-term survival benefits.
• Targeting the BRAFV600E mutation and MAPK pathway has shown promise, but response rates remain suboptimal.

Purpose of the Study:

• To identify key gene expression and copy number features that stratify melanoma cell lines.
• To develop a machine learning approach for selecting informative genes from high-dimensional genomic data.
• To discover novel biomarkers for predicting melanoma subtypes and potential treatment responses.

Main Methods:

• Utilized the Cancer Cell Line Encyclopedia (CCLE) dataset with 62 melanoma samples.
• Performed clustering analysis on gene expression (19K+) and copy number (20K+) data.
• Developed an integrated machine learning pipeline for high-dimensional feature selection.

Main Results:

• A small subset of 15 genes was sufficient for accurate classification of melanoma cell line clusters.
• The selected genes achieved near-perfect performance in test split classifications.
• Identified known melanoma-associated genes and novel candidates for further investigation.

Conclusions:

• A concise set of genomic features can effectively distinguish melanoma cell line subtypes.
• This gene signature holds potential for developing predictive biomarkers in metastatic melanoma.
• Further research into the novel genes identified could uncover new therapeutic targets.