model.interpretation.plot: Model Interpretation Plot

In SIAMCAT: Statistical Inference of Associations between Microbial Communities And host phenoTypes

Description

This function produces a plot for model interpretation, displaying

• the feature weights,

• the robustness of feature weights

• the features scores across samples,

• the distribution of metadata across samples, and

• the proportion of model weights shown.

Usage

model.interpretation.plot(siamcat, fn.plot = NULL,
    color.scheme = "BrBG",
    consens.thres = 0.5,
    heatmap.type = "zscore",
    limits = c(-3, 3), detect.lim = 1e-06,
    max.show = 50, prompt=TRUE, verbose = 1)

Arguments

siamcat	object of class siamcat-class
fn.plot	string, filename for the pdf-plot
color.scheme	color scheme for the heatmap, defaults to 'BrBG'
consens.thres	float, minimal ratio of models incorporating a feature in order to include it into the heatmap, defaults to 0.5 Note that for 'randomForest' models, this cutoff specifies the minimum median Gini coefficient for a feature to be included and should therefore be much lower, e.g. 0.01
heatmap.type	string, type of the heatmap, can be either 'fc' or 'zscore', defaults to 'zscore'
limits	vector, cutoff for extreme values in the heatmap, defaults to c(-3, 3)
detect.lim	float, pseudocount to be added before log-transformation of features, defaults to 1e-06
max.show	integer, maximum number of features to be shown in the model interpretation plot, defaults to 50
prompt	boolean, turn on/off prompting user input when not plotting into a pdf-file, defaults to TRUE
verbose	control output: 0 for no output at all, 1 for only information about progress and success, 2 for normal level of information and 3 for full debug information, defaults to 1

Details

Produces a plot consisting of

• a barplot showing the feature weights and their robustness (i.e. in what proportion of models have they been incorporated)

• a heatmap showing the z-scores of the metagenomic features across patients

• another heatmap displaying the metadata categories (if applicable)

• a boxplot displaying the poportion of weight per model that is actually shown for the features that are incorporated into more than consens.thres percent of the models.

Value

Does not return anything, but produces the model interpretion plot.

Examples

data(siamcat_example)

# simple working example
model.interpretation.plot(siamcat_example, fn.plot='./interpretion.pdf',
    heatmap.type='zscore')

SIAMCAT documentation built on Nov. 8, 2020, 5:14 p.m.