Towards Improving Skin Cancer Diagnosis by Integrating Microarray and RNA-Seq Datasets

Area of Science:

• Dermatology
• Genomics
• Bioinformatics

Background:

• Biological similarities among skin conditions complicate accurate diagnosis.
• Gene expression analysis offers a promising avenue for identifying biomarkers.
• Developing intelligent clinical decision support systems is crucial for improved diagnostics.

Purpose of the Study:

• To present a novel computational approach for integrating heterogeneous transcriptomic data.
• To identify a panel of differentially expressed genes (DEGs) for discerning multiple skin pathologies.
• To enhance the accuracy of skin cancer diagnosis through advanced bioinformatics.

Main Methods:

• Integration of multiple heterogeneous transcriptomic datasets.
• Development of pipelines for batch merging, biomarker selection, and classification assessment.
• Utilizing a panel of 8 multiclass DEGs to classify 10 distinct skin states.

Main Results:

• A panel of 8 highly relevant multiclass DEGs was identified for distinguishing up to 10 skin pathological states.
• Classification models achieved high diagnostic power, with overall and mean multiclass F1-scores exceeding 94% and 80%, respectively.
• The study demonstrated robust performance in diagnosing new samples using the identified gene panel.

Conclusions:

• The identified 8 multiclass DEGs are highly relevant for differentiating various skin conditions, including cancerous states.
• The developed approach provides new insights for improving skin cancer diagnosis.
• Clinicians should consider the biological significance of these DEGs for clinical application.