Domain-Scan: Combinatorial Sero-Diagnosis of Infectious Diseases Using Machine Learning

Area of Science:

• Immunology
• Bioinformatics
• Infectious Diseases

Background:

• Antibodies in blood indicate past pathogen exposure, typically measured by immunoassays.
• Current methods like ELISA and western blots have limitations in high-throughput antibody profiling.

Purpose of the Study:

• To introduce Domain-Scan, a new sero-diagnostic tool for detecting pathogen-specific antibodies.
• To demonstrate the utility of Domain-Scan for classifying infections using machine learning.

Main Methods:

• Phage-display epitope arrays were used to create "domains" representing pathogen epitopes.
• Next-generation sequencing (NGS) quantified serum antibody binding to multiple epitopes simultaneously.
• Machine learning models were trained to classify samples based on antibody binding profiles.

Main Results:

• Domain-Scan accurately classified individuals infected with HIV-1 or HCV.
• The method identified specific "domains" (epitopes) crucial for accurate classification.
• High-throughput antibody profiling was achieved by measuring binding to dozens of epitopes.

Conclusions:

• Domain-Scan offers a powerful and generalizable approach for sero-diagnosis.
• The integration of phage-display, NGS, and machine learning enables precise antibody detection.
• This method can be adapted for diagnosing infections by various pathogens with sufficient training data.