Skip to main content

Table 2 Overview of Evaluated Methods

From: Pathway activity inference for multiclass disease classification through a mathematical programming optimisation framework

Guo et al. [28]

Abbreviation: Mean

Computational basis: Pathway activity

Description: Create pathway-specific gene expression profiles; for each pathway, pathway activity for sample is its mean expression value among all member genes; a classifier is trained on pathway activity profile.

Guo et al. [28]

Abbreviation: Median

Computational basis: Pathway activity

Description: Create pathway-specific gene expression profiles; for each pathway, pathway activity for sample is its median expression value among all member genes; a classifier is trained on pathway activity profile.

Bild et al. [45]

Abbreviation: PCA

Computational basis: Pathway activity

Description: Create pathway-specific gene expression profiles; for each pathway, top principal component is calculated as the pathway activity; a classifier is trained on pathway activity profile.

Lee et al. [19]

Abbreviation: CORGs

Computational basis: Pathway activity

Description: Create pathway-specific gene expression profiles; for each pathway, apply t-test to rank genes and perform a greedy search to find a subset of genes whose averaged expression values is locally maximal in t-test value; a classifier is trained on pathway activity profile; only applicable for two-class problems.

Ainali et al. [72]

Abbreviation: Per_pathway

Computational basis: Single genes

Description: Create pathway-specific gene expression profiles; a classifier is trained on each pathway-specific gene expression profile separately, and prediction rates achieved by all pathway classifiers are averaged as the final prediction rate.

Single Genes

Abbreviation: SG

Computational basis: Single genes

Description: Apply [71] to select a subset of top genes; a classifier is trained on reduced gene expression profile

Proposed in this work

Abbreviation: DIGS

Computational basis: Pathway activity

Description: Create pathway-specific gene expression profiles; Apply the proposed DIGS model to construct pathway activity as weighted linear summation of gene expressions; a classifier is trained on pathway activity profile.