Data-Independent Acquisition Mass Spectrometry of EPS-Urine Coupled to Machine Learning: A Predictive Model for...

Area of Science:

• Urology
• Proteomics
• Biomarker Discovery

Background:

• Prostate cancer (PCa) diagnosis relies on prostate-specific antigen (PSA) testing and digital rectal exams (DRE), which lack specificity and sensitivity.
• Current methods struggle to differentiate aggressive from indolent PCa, necessitating improved diagnostic approaches and novel biomarkers.
• Expressed prostatic secretion (EPS)-urine analysis offers a promising avenue for non-invasive biomarker discovery.

Purpose of the Study:

• To identify differentially expressed proteins in EPS-urine samples from PCa and benign prostatic hyperplasia (BPH) patients.
• To develop a predictive model for PCa detection using proteomic data and clinical parameters.
• To enhance the accuracy and specificity of PCa diagnosis beyond current screening methods.

Main Methods:

• EPS-urine samples from 133 patients (PCa and BPH) were analyzed using data-independent acquisition (DIA) mass spectrometry for high-sensitivity proteomic profiling.
• A total of 2615 proteins were identified, with 1670 consistently detected across all samples.
• Machine learning algorithms were employed to build a predictive model integrating protein quantification with clinical data (PSA level, gland size).

Main Results:

• The best predictive model incorporated semaphorin-7A (sema7A), secreted protein acidic and rich in cysteine (SPARC), FT ratio, and prostate gland size.
• This model achieved 83% accuracy in distinguishing between PCa and BPH in the validation set.
• The study achieved the highest proteomic coverage to date for EPS-urine samples.

Conclusions:

• Proteomic analysis of EPS-urine combined with machine learning can significantly improve PCa diagnostic accuracy.
• Semaphorin-7A and SPARC are identified as potential novel biomarkers for PCa detection.
• This approach offers a more specific and sensitive method for differentiating PCa from BPH.