Immune subtype identification and multi-layer perceptron classifier construction for breast cancer

Area of Science:

• Oncology
• Immunology
• Bioinformatics

Background:

• Breast cancer is a complex disease influenced by its tumor microenvironment (TME).
• Understanding tumor-infiltrating immune cells (TIICs) is crucial for predicting treatment response and prognosis.

Purpose of the Study:

• To classify breast cancer subtypes based on TIIC composition.
• To develop a deep learning model for accurate breast cancer subtype diagnosis.

Main Methods:

• Consensus clustering and K-means algorithms were used to group breast cancer subtypes.
• A multi-layer perceptron (MLP) classifier was developed using a 38-gene signature.
• TIIC proportions, immune checkpoint molecule expression, and tumor mutational burden (TMB) were analyzed.

Main Results:

• Two distinct breast cancer subtypes were identified with significant differences in overall survival (OS).
• Subtypes exhibited varying proportions of key immune cells (e.g., CD8, CD4, regulatory T cells).
• A 38-gene signature-based MLP classifier achieved 93.56% accuracy in subtype diagnosis.

Conclusions:

• Immune signatures in the TME can effectively differentiate breast cancer subtypes.
• Accurate subtype identification can guide personalized treatment strategies and improve patient outcomes.