Pre-trained knowledge transfer with application in multisystem inflammatory syndrome subphenotyping in children

Area of Science:

• Pediatric infectious diseases
• Computational biology
• Epidemiology

Background:

• Multisystem Inflammatory Syndrome in Children (MIS-C) is a severe post-SARS-CoV-2 condition with diverse clinical presentations.
• Identifying MIS-C subphenotypes is crucial for risk stratification and targeted therapies.
• Analyzing combined data from multiple studies is ideal but hindered by privacy and heterogeneity.

Purpose of the Study:

• To develop a novel framework for transferring knowledge of MIS-C subphenotypes from large source studies to a target study.
• To overcome challenges of patient-level data sharing and between-study heterogeneity in MIS-C research.
• To enable robust subphenotype characterization even with limited target study data.

Main Methods:

• A one-shot summary statistics-based framework utilizing pre-trained subphenotype knowledge.
• Incorporation of study-specific subphenotype mixing proportions to account for heterogeneity.
• Validation through real-data guided simulation studies and application to MIS-C data.

Main Results:

• The proposed framework successfully transfers knowledge of shared MIS-C subphenotypes.
• Study-specific mixing proportions effectively explain between-study clinical heterogeneity.
• The method demonstrates significant benefits in analyzing MIS-C data compared to traditional approaches.

Conclusions:

• The developed framework provides an effective solution for joint analysis of multi-source MIS-C data.
• It facilitates robust subphenotype identification while preserving data privacy and managing heterogeneity.
• This approach can advance understanding and treatment strategies for MIS-C.