post: Phylogeney-Guided OTU-Specific Association Test for...

In POSTm: Phylogeny-Guided OTU-Specific Association Test for Microbiome Data

Phylogeney-Guided OTU-Specific Association Test for Microbiome Data

Description

The POST implements a phylogeny-guided OTU-specific association test for microbiome data. This method boosts the testing power by adaptively borrowing information from phylogenetically close OTUs of the target OTU. Whether or not borrowing information or the amount of information from the neighboring OTUs is data adaptive and supervised by phylogenetic distance and the outcome variable. POST is built on a kernel machine regression framework and inherited the advantages including flexibly model complex microbiome effects (e.g., effects from opposite direction), easily adjust for covariates, and accommodate both continuous and binary outcomes.

Usage

post(
  y,
  OTU,
  tree = NULL,
  X = NULL,
  cValues = seq(from = 0, to = 0.05, by = 0.01)
)

Arguments

y	A numerical vector. The outcome of interest. Data can be binary or continuous.
OTU	A matrix object. The operational taxonomic units (OTU). Data can be provided as counts or as proportions. Each row indicates a single sample; each column a single OTU. NA/0 values are allowed, but their presence will trigger a shift of all data by a small internally defined value. The matrix must include column headers providing unique identification for each OTU; these identifiers are expected to be included in the tip labels of the input tree object. Any identifiers that are not included as tip labels are removed from the analysis.
tree	An object of class "phylo", "hclust", "phylog", a matrix object or NULL. If NULL, only the single OTU test will be estimated. Objects of class "phylo", "hclust", and "phylog" are phylogenetic trees, the tip labels of which must include all of the identifiers used as column headers of OTU. If a matrix, a square symmetric matrix containing the pairwise distances between OTUs as defined by the branch lengths. Note that the full tree should be provided/used and should not be subset or truncated, even if OTU does not contain all tips. See details for further information.
X	A data.frame object, matrix object or NULL. The covariates data. If NULL, an intercept only model is assumed. Factor covariates are allowed.
cValues	A numeric vector. The c values at which p-values are to be estimated. The default is a vector of evenly spaced values between zero and the recommended maximum value for OTUs defined at 97% sequence similiary, c_max = 0.05. If no tree is provided, cValues will be set to 0.

Details

It is assumed that the OTU table is defined by a 97% sequence similarity. Though this threshold is not (cannot be) enforced in the implementation, the recommended maximum c-value may not be appropriate for other thresholds.

There are numerous packages available for generating phylogenetic trees. The sole purpose of the tree input is to enable the calculation of pairwise distances between the pairs of tips using branch lengths. Because it is not feasible to support this functionality for all packages that generate such trees, the package allows for the specification of the distance matrix as an alternative to providing a specifically formatted tree object. For objects of class "phylo", "hclust", and "phylog", the ape package provides tools to obtain the distance matrix, and these tools are used in this implementation. For all others, the distance matrix (defined by branch lengths) must be provided by the user through input tree.

Value

Returns a POST object. Analysis results are provided as a matrix, the contents of which will depend on the inputs selected for the analysis. Possible columns include

POST_pvalue	A numeric object. The POST p-value.
SO_pvalue	A numeric object. The single OTU test p-value.
BEST_C	A numeric object. The c value corresponding to the minimum POST p-value.

Row names indicate the OTU to which the results pertain. Results are ordered according to the POST (if tree provided) or SO (if tree not provided) p-values.

References

Huang, C., Callahan, B., Wu, M. C., Holloway, S. T., Brochu, H., Lu, W., Peng, X., and Tzeng, J-Y. (2021). Phylogeny-guided microbiome OTU-specific association test (POST). Bioinformatics, under revision.

Examples

data("POSTmData")

y <- as.integer(x = metadata[,"GC"] == "BV")
X <- metadata[,"mRace"]

result <- post(y = y, 
               X = X, 
               OTU = otu[,1:20], 
               tree = otutree,
               cValues = seq(0,0.05,by=0.01))

POSTm documentation built on May 29, 2024, 9:24 a.m.