GeneWalk identifies relevant gene functions for a biological context using network representation learning

Table 1 Overview of GeneWalk and alternative methods used for systematic comparison of gene function relevance scoring. The alternative methods were selected based on prevalence of usage or characteristic model features

		Requirements
Method	Objective	Input type	GO annotations	Gene network	Defining model characteristic
GeneWalk	Gene function relevance scoring	Gene list	Yes	Yes	Network representation learning (cosine similarity)
PANTHER	Gene set enrichment	Gene list	Yes	No	Overrepresentation analysis (Fisher Exact test)
GeneMANIA	Gene function prediction (binary classification)	Gene list	Yes	Yes	Network label propagation algorithm
GGEA	Gene set enrichment	Quantitative expression score for all genes	Yes	Yes	Gene set overrepresentation analysis accounting for gene network
GSEA	Gene set enrichment	Quantitative expression score for all genes	Yes	No	Gene set enrichment analysis (permutation score test)
MGSA	Gene set relevance scoring	Gene list	Yes	No	Bayesian network (posterior probability)
PADOG	Gene set enrichment	Expression levels for all genes	Yes	No	Pathway Analysis with Down-weighting of Overlapping Genes (permutation score test)
STRING	Gene set enrichment	Gene list	Yes	No	Overrepresentation analysis (Hypergeometric test)
topGO	Gene set enrichment	Gene list	Yes	No	Overrepresentation analysis (Fisher Exact test) with decorrelation of parental GO terms

ISSN: 1474-760X