Development and Validation of a Machine-Learning Deep Plasma Proteome Classifier for Early-Stage Breast Cancer...

Area of Science:

• Proteomics
• Biomarker Discovery
• Early Cancer Detection

Background:

• Liquid biopsy tests analyzing plasma proteomes offer potential for early cancer detection.
• Breast cancer diagnosis often relies on imaging and invasive procedures, highlighting the need for non-invasive early detection methods.

Purpose of the Study:

• To evaluate the accuracy and utility of a proteomic-based liquid biopsy test for early breast cancer detection.
• To assess the performance of the test as a supplemental tool to mammography.

Main Methods:

• Analysis of plasma proteomes from 1,259 women (healthy and with breast cancer) using label-free mass spectrometry.
• Development and validation of a machine learning classifier trained on a cohort of 845 women and validated on 397 women.
• Held-out validation was performed using automated, blinded sample processing.

Main Results:

• The test achieved 92.6% sensitivity and 92.3% specificity in held-out validation, with an AUC of 0.975.
• High sensitivity was observed across various breast cancer stages and subtypes.
• Simulated data indicated the test could detect 93% of mammography-missed cancers and significantly reduce false positives compared to other imaging methods.

Conclusions:

• Proteomic analysis of plasma demonstrates high accuracy for breast cancer detection, particularly in early stages.
• The developed test is a promising supplemental screening tool for women, especially when used in conjunction with mammography.
• Further research into identified pathways like EMT and PI3K-AKT signaling may enhance understanding of early breast cancer biology.