PROMISE

PROMISE identifies genomic variables that exhibit biologically defined multivariate association patterns across multiple endpoint variables by projecting observed association statistics onto a vector representing the most interesting pattern.

Key Features:

• Biological Knowledge Integration: Constructs a vector encoding biologically significant target values for statistics that characterize associations between genomic variables and multiple endpoints.
• Test Statistic (Dot-product): Computes a test statistic as the dot product of the observed association statistics vector and the most-interesting-values vector to quantify alignment with the predefined pattern.
• Projection-Based Measure: Interprets the test statistic as the length of the projection of the observed correlation vector onto the vector of most interesting associations to measure pattern strength.
• Statistical Significance Assessment: Assesses significance using permutation testing.

Scientific Applications:

• Multi-endpoint genomic association discovery: Identifies genomic features that follow specific patterns of association across multiple phenotypic or clinical endpoints.
• Comparative power advantage: Offers greater statistical power relative to classical multivariate procedures, individual endpoint analyses, or methods that only list genes significant across endpoints without considering association patterns.

Methodology:

Constructs a vector of biologically significant values for association statistics, computes the dot-product between the observed association-statistics vector and that vector (interpreted as the projection length of the observed correlation vector), and evaluates significance by permutation testing.