createTIEC: createTIEC

In mcalgaro93/benchdamic: Benchmark of differential abundance methods on microbiome data

createTIEC

Description

Extract the list of p-values from the outputs of the differential abundance detection methods to compute several statistics to study the ability to control the type I error and the p-values distribution.

Usage

createTIEC(object)

Arguments

object

Output of the differential abundance tests on mock comparisons. Must follow a specific structure with comparison, method, matrix of p-values, and method's name (See vignette for detailed information).

Value

A list of data.frames:

• df_pval 5 columns per number_of_features x methods x comparisons rows data.frame. The four columns are called Comparison, Method, variable (containing the feature names), pval, and padj;

• df_FPR 5 columns per methods x comparisons rows data.frame. For each set of method and comparison, the proportion of false positives, considering 3 thresholds (0.01, 0.05, 0.1) are reported;

• df_FDR 4 columns per methods rows data.frame. For each method, the average proportion of mock comparisons where false positives are found, considering 3 thresholds (0.01, 0.05, 0.1), are reported. Each value is an estimate of the nominal False Discovery Rate (FDR);

• df_QQ contains the coordinates to draw the QQ-plot to compare the mean observed p-value distribution across comparisons, with the theoretical uniform distribution;

• df_KS 5 columns and methods x comparisons rows data.frame. For each set of method and comparison, the Kolmogorov-Smirnov test statistics and p-values are reported in KS and KS_pval columns respectively.

See Also

createMocks

Examples

# Load some data
data(ps_stool_16S)

# Generate the patterns for 10 mock comparison for an experiment
# (N = 1000 is suggested)
mocks <- createMocks(nsamples = phyloseq::nsamples(ps_stool_16S), N = 10)
head(mocks)

# Add some normalization/scaling factors to the phyloseq object
my_norm <- setNormalizations(fun = c("norm_edgeR", "norm_CSS"),
    method = c("TMM", "CSS"))
ps_stool_16S <- runNormalizations(normalization_list = my_norm,
    object = ps_stool_16S)

# Initialize some limma based methods
my_limma <- set_limma(design = ~ group, coef = 2,
    norm = c("TMM", "CSS"))

# Run methods on mock datasets
results <- runMocks(mocks = mocks, method_list = my_limma,
    object = ps_stool_16S)

# Prepare results for Type I Error Control
TIEC_summary <- createTIEC(results)

# Plot the results
plotFPR(df_FPR = TIEC_summary$df_FPR)
plotFDR(df_FDR = TIEC_summary$df_FDR)
plotQQ(df_QQ = TIEC_summary$df_QQ, zoom = c(0, 0.1))
plotKS(df_KS = TIEC_summary$df_KS)
plotLogP(df_QQ = TIEC_summary$df_QQ)

mcalgaro93/benchdamic documentation built on Nov. 28, 2024, 2:16 p.m.