Disease diagnostics using machine learning of B cell and T cell receptor sequences

Area of Science:

• Immunology
• Computational Biology
• Medical Diagnostics

Background:

• Current clinical diagnosis relies on physical exams, history, lab tests, and imaging, often overlooking the immune system's memory of antigen exposures.
• B cell and T cell receptors encode a detailed history of an individual's encounters with antigens, representing a largely untapped diagnostic resource.

Purpose of the Study:

• To develop an interpretive framework, MAchine Learning for Immunological Diagnosis (MILAD), for simultaneous or precise screening of multiple illnesses.
• To leverage immune receptor repertoire data for enhanced disease detection and characterization.

Main Methods:

• Analysis of immune receptor datasets from 593 individuals.
• Development of a machine learning model for immunological diagnosis.
• Validation of model interpretability by comparing with known immune responses.

Main Results:

• The MILAD framework successfully screens for multiple illnesses, including specific infections (e.g., SARS-CoV-2, influenza, HIV), autoimmune disorders (e.g., lupus, type-1 diabetes), and vaccine responses.
• The model identifies distinct immunological signatures for different diseases and quantifies disease severity differences.
• Interpretable features of the model align with established immunological knowledge and highlight antigen-specific receptors.

Conclusions:

• MAchine Learning for Immunological Diagnosis offers a novel approach to clinical diagnostics by utilizing the immune system's receptor data.
• This framework has broad potential for the scientific and clinical interpretation of immune responses, enabling simultaneous or targeted disease screening.
• MILAD enhances diagnostic capabilities by uncovering specific infections, autoimmune conditions, and vaccine efficacies through immune receptor profiling.