MTOP

MTOP implements a mixed-effect two-stage polytomous model score test to evaluate associations between genetic variants and cancer subtypes by incorporating multivariate tumor characteristics into case-control odds ratio estimation.

Key Features:

• Two-Stage Polytomous Regression Framework: Stage one uses a polytomous regression model to define cross-classified cancer subtypes based on histopathological features and molecular tumor markers.
• Parsimonious Modeling: Stage two refines subtype-specific case-control odds ratios via a parsimonious mixed-effect model that parameterizes a baseline subtype odds ratio and parameters associated with tumor markers.
• Handling of Correlated Tumor Features and Missing Data: An Expectation-Maximization algorithm is used to manage correlated tumor features and missing tumor marker data.
• Degrees-of-Freedom Reduction via Random-Effect Modeling: Additional case-case parameters for exploratory markers can be specified as random effects to reduce degrees-of-freedom.
• Simulation and Real-World Validation: Performance has been evaluated through extensive simulations and validated using data from the Polish Breast Cancer Study (PBCS).

Scientific Applications:

• Genome-wide association studies (GWAS): Enables testing genetic associations across heterogeneous cancer subtypes by integrating multivariate tumor characteristics.
• Subtype-specific association discovery: Identifies and characterizes heterogeneous associations between genetic loci and distinct tumor subtypes.

Methodology:

Polytomous regression to classify cross-classified tumor subtypes; a mixed-effect two-stage polytomous model score test (MTOP) to refine subtype-specific case-control odds ratios using fixed and random effects; an Expectation-Maximization algorithm to handle correlated tumor features and missing data; specification of random-effect case-case parameters to reduce degrees-of-freedom; and assessment via simulation studies and validation with the Polish Breast Cancer Study (PBCS).