GMPR {GUniFrac}

Geometric Mean of Pairwise Ratios (GMPR) Normalization for Zero-inflated Count Data

Description

A robust normalization method for zero-inflated count data such as microbiome sequencing data.

Usage

GMPR(OTUmatrix, min_ct = 2, intersect_no = 4)

Arguments

OTUmatrix

An OTU count table, where OTUs are arranged in rows and samples in columns.

min_ct

The minimal number of OTU counts. Only those OTU pairs with at least min_ct counts are considered in the ratio calculation. The default is 2.

intersect_no

The minimal number of shared OTUs between samples. Only those sample pairs sharing at least intersect_no OTUs are considered in geometric mean calculation. The default is 4.

Details

Normalization is a critical step in microbiome sequencing data analysis to account for variable library sizes. Microbiome data contains a vast number of zeros, which makes the traditional RNA-Seq normalization methods unstable. The proposed GMPR normalization remedies this problem by switching the two steps in DESeq2 normalization:

First, to calculate rij, the median count ratio of nonzero counts between samples: rij=median(cki/ckj) (k in 1:OTU_number and cki, ckj is the non-zero count of the kth OTU)

Second, to calculate the size factor si for a given sample i: si=geometric_mean(rij)

Value

A vector of GMPR size factor for each sample.

Author(s)

Jun Chen and Lujun Zhang

References

Li Chen, James Reeve, Lujun Zhang, Shenbing Huang, and Jun Chen. 2018. GMPR: A robust normalization method for zero-inflated count data with application to microbiome sequencing data. PeerJ, 6, e4600.

Examples

data(throat.otu.tab)
size.factor <- GMPR(t(throat.otu.tab))