OMiSALN: Otimal Microbiome-based Survival Analysis using Linear and...

In hk1785/OMiSA: Optimal Microbiome-based Survival Analysis (OMiSA)

Description

OMiSALN is a non-parametric method which tests the association between a microbial group (e.g., community, taxon) composition and a survival (time-to-event) response on human health or disease with or without covariate adjustments (e.g., age, sex).

Usage

OMiSALN(obstime, delta, X, total.reads = NULL, cov = NULL, 
pow = c(1/4, 1/3, 1/2, 1), n.perm = 5000)

Arguments

obstime	A numeric vector for the observed times.
delta	A numeric vector for the event/censoring indicators (1: event, 0: censoring).
X	A matrix for the OTU table (1. Elements are counts. 2. Rows are samples and columns are OTUs. 3. Monotone/singletone OTUs must be removed.).
total.reads	A numeric vector for the total reads per sample in the entire community. If you survey the entire community, you do not need to specify this. If you test a microbial taxon, you need to specify this (see the examples below). Default is NULL for the entire community.
cov	A data.frame (or vector) for covariate adjustment(s) (Rows are samples and columns are covariate variables).
pow	A set of the candidate gamma values. Default is c(1/4, 1/3, 1/2, 1).
n.perm	The number of permutations. Default is 5000.

Value

pvs.misaln: The estimated p-values for individual MiSALN tests

p.omisaln: The estimated p-value for OMiSALN

Author(s)

Hyunwook Koh

References

Koh H, Livanos AE, Blaser MJ, Li H. (2018) A highly adaptive microbiome-based association test for survival traits. BMC Genomics 19, 210

Examples

library(dirmult) 
library(phyloseq) 
library(robustbase)
library(robCompositions) 
library(BiasedUrn)
library(CompQuadForm)
library(GUniFrac) 
library(ecodist) 
library(survival)
library(OMiSA)

### An example simulated microbiome data with survival outcomes

data(MiSurv.Data) 

otu.tab <- otu_table(MiSurv.Data)
tax.tab <- tax_table(MiSurv.Data)
obstime <- as.numeric(unlist(sample_data(MiSurv.Data)[,1]))
delta <- as.numeric(unlist(sample_data(MiSurv.Data)[,2]))
x1 <- as.numeric(unlist(sample_data(MiSurv.Data)[,3]))
x2 <- as.numeric(unlist(sample_data(MiSurv.Data)[,4]))

covs <- as.data.frame(cbind(x1, x2))
covs[,2] <- as.factor(covs[,2])

### Example 1. To test the entire community (e.g., kingdom)

set.seed(100)

OMiSALN(obstime, delta, otu.tab, total.reads=NULL, cov=covs)

### Example 2. To test the higher-level taxon, p__Firmicutes

# Important notice: Create n total reads counts "in the entire community" and 
# specify it into the element, "total.reads", of the function, OMiSALN, as below.
total.reads <- rowSums(otu.tab)

ind.Firmicutes <- which(tax.tab[,2] == "p__Firmicutes")
otu.tab.Firmicutes <- otu.tab[,ind.Firmicutes]

set.seed(100)

OMiSALN(obstime, delta, otu.tab.Firmicutes, total.reads=total.reads, cov=covs)

### Example 3. To test the higher-level taxon, p__Bacteroidetes

# Important notice: Create n total reads counts "in the entire community" and 
# specify it into the element, "total.reads", of the function, OMiSALN, as below.
total.reads <- rowSums(otu.tab)

ind.Bacteroidetes <- which(tax.tab[,2] == "p__Bacteroidetes")
otu.tab.Bacteroidetes <- otu.tab[,ind.Bacteroidetes]

set.seed(100)

OMiSALN(obstime, delta, otu.tab.Bacteroidetes, total.reads=total.reads, cov=covs)

hk1785/OMiSA documentation built on March 1, 2020, 6:38 p.m.