stepwiseCM

stepwiseCM classifies cancer samples using two heterogeneous datasets while preserving each dataset's distinct classification power.

Key Features:

• Two-Stage Classification Strategy: It operates in two stages, first using a dataset with full measurements for all samples (e.g., clinical covariates) and then incorporating a second dataset with partial, high-dimensional measurements (e.g., gene expression).
• Cost-Efficiency: It minimizes additional expensive or invasive measurements by identifying a subgroup of samples that are most likely to benefit from obtaining the second set of covariates.
• Indirect Mapping Between Data Spaces: It connects heterogeneous data spaces via indirect mapping by projecting neighborhood information from one data space onto another, enabling integration of high-dimensional types such as DNA copy number and mRNA expression.
• Rapid Diagnosis and Reduced Patient Distress: By avoiding unnecessary measurements for many individuals, it can shorten time to result and reduce patient distress.

Scientific Applications:

• Cancer Sample Classification: Classifying cancer samples using diverse datasets to leverage complementary predictive signals while avoiding direct merging.
• Integration of High-Dimensional Data Types: Pairing high-dimensional datasets such as DNA copy number and mRNA expression for joint classification through indirect mapping.

Methodology:

It uses a stepwise classification strategy that preserves each dataset's distinct classification power without merging them and projects neighborhood information from one data space onto another to identify samples that should receive additional measurements.