kruskal_abundance: Kruskal-Wallis differential abundance analysis

In umerijaz/microbiomeSeq: Microbial community analysis in an environmental context.

Description

This function finds the features that are significantly differentially abundant in the provided taxa abundance data under different conditions using Kruskal-Wallis test. The p-values values generated are corrected for multiple testing using family wise error rate. Significance is based on the corrected pvalue threshold. Significant features are assigned importance using random forest classifier. The measure of importance used in this case is mean decrese accuracy.

Usage

kruskal_abundance(physeq, grouping_column, pvalue.threshold = 0.05)

Arguments

physeq	(Required). A phyloseq object containing merged information of abundance, taxonomic assignment, sample data including the measured variables and categorical information of the samples, and / or phylogenetic tree if available.
grouping_column	(Required). Character string specifying name of a categorical variable that is preffered for grouping the information. information, this should be one of the components of grouping vector.
pvalue.threshold.	Cut off p-value for significance of differentially abundant taxa, default is 0.05.

Value

Returns a list of three items:

• SignfeaturesTable: A data.frame of taxa with coresponding raw p-values, corrected p-values, family wise error rate and expected abundance computed by using raw p-values.

• importance: A data.frame of taxa mean decrease in accuracy as obtained by random forest classifier.

• plotdata: A data.frame of taxa and corresponding corrected p-values, importance rank organised in form accepted by ggplot.

Author(s)

Alfred Ssekagiri [e-mail], Umer Zeeshan Ijaz [e-mail]

References

http://userweb.eng.gla.ac.uk/umer.ijaz/, Umer Ijaz, 2015

http://www.bigre.ulb.ac.be/courses/statistics_bioinformatics/practicals/microarrays_berry_2010/berry_feature_selection.html

Examples

data(pitlatrine)
physeq <- pitlatrine
physeq <- taxa_level(physeq,"Phylum")
kw_sig <-  kruskal_abundance(physeq, "Country")
#plot the significant features
plot_signif(kw_sig$plotdata) #see function \link[microbiomeSeq]{plot_signif}

umerijaz/microbiomeSeq documentation built on May 30, 2019, 3:13 p.m.