Lightweight, open source, easy-use algorithm and web service for paraprotein screening using spatial frequency domain...

Area of Science:

• Clinical Chemistry
• Computational Biology
• Oncology

Background:

• Serum protein electrophoresis (SPEP) is a standard diagnostic tool for plasma cell neoplasms.
• Accurate detection of monoclonal paraproteins is crucial for laboratory diagnosis.
• Current methods may require significant computational resources.

Purpose of the Study:

• To develop and validate an automated, computationally efficient screening method for paraprotein detection using SPEP.
• To create accessible tools for paraprotein screening in clinical laboratories.

Main Methods:

• A model was developed based on high-frequency components in the spatial frequency spectrum of SPEP densitometry curves.
• The model was trained on 330 samples and validated on representative (n=110) and external (n=1,321) test sets.
• An interactive web application and a macro-enabled spreadsheet were created for user-friendly screening.

Main Results:

• The model achieved an Area Under the Curve (AUC) of 0.90 in both representative and external test sets.
• Sensitivity for paraprotein detection was high, reaching 0.97 in the external test set.
• The method demonstrated effective detection across low, medium, and high paraprotein concentrations.

Conclusions:

• An automated, computationally efficient method for paraprotein screening using SPEP curve characteristics has been successfully developed.
• The developed tools, including a web service and open-source package, facilitate widespread adoption and customization.
• Future research should explore integration with other laboratory data and clinical information.