GFA

GFA identifies latent factors across clinical, cytokine, genomic, methylation, and dietary datasets to reveal multi-level molecular mechanisms underlying conditions such as obesity.

Key Features:

• Integrative Analysis: Jointly analyzes clinical, cytokine, genomic, methylation, and dietary data to detect shared and view-specific latent factors.
• Application to Twin Studies: Applied to monozygotic twin pairs with weight discordance (delta BMI > 3 kg/m²) to distinguish genetic and environmental contributions to obesity.
• Handling Large Datasets: Manages extensive datasets as demonstrated on TwinFat study data comprising 42, 71, 1587, 1605, and 63 variables respectively.
• Hypothesis Generation: Facilitates generation of hypotheses linking cytokines, dietary habits, inflammatory responses, and epigenetic (methylation) modifications to weight gain and related phenotypes.

Scientific Applications:

• Obesity mechanism discovery: Dissects multi-level molecular mechanisms in obesity using integrated multi-omics and clinical data from twin studies.
• Complex disease integrative analysis: Integrates diverse molecular and clinical datasets to investigate diseases involving multiple biological pathways and to inform identification of therapeutic targets and patient stratification.

Methodology:

GFA employs a joint integrative machine learning framework that simultaneously analyzes multiple data modalities to identify latent factors representing underlying biological processes.