SCFA

SCFA removes noisy signals from consistent molecular patterns in multi-omics data to identify cancer subtypes and predict patient risk scores.

Key Features:

• Noise reduction: Efficiently removes noisy signals while preserving consistent molecular patterns across multi-omics data.
• Feature selection: Uses an autoencoder to select only important features from the input data.
• Factor analysis: Repeatedly performs factor analysis to represent data with different numbers of factors.
• Cancer subtyping: Reliably identifies cancer subtypes from the factor representations.
• Risk prediction: Accurately predicts risk scores of patients using the derived representations.

Scientific Applications:

• Cancer subtyping: Identifies molecular subtypes of cancer from multi-omics data integration.
• Prognostic risk stratification: Predicts patient risk scores for clinical outcome assessment.
• Multi-omics integration: Analyzes consistent patterns across multiple molecular data types.

Methodology:

SCFA uses an autoencoder for feature selection, then repeatedly performs factor analysis with varying numbers of factors to generate data representations for downstream subtyping and risk prediction.