JBA_binning: Binning of pJRES spectra

In MWASTools: MWASTools: an integrated pipeline to perform metabolome-wide association studies

Description

This function performs binning of pJRES spectra using the JBA algorithm (see details).

Usage

JBA_binning (NMR_data, st = 4, ct = 0.85, int = "sum", cm = "pearson",
             ef = 2, merge = TRUE, mt = 0.9)

Arguments

NMR_data	numeric matrix containing the NMR data (i.e. NMR peak intensities). The columns of the matrix must correspond to the metabolic variables (chemical shifts) and the rows to the samples. Column and row names must contain the metabolite IDs (i.e chemical shifts) and the sample IDs, respectively.
st	numeric value indicating the minimum bin size.
ct	numeric value indicating the correlation threshold. Bins with average correlation below ct will be neglected. This value can be established by comparing the distribution of average correlations in a spectral region dominated by electronic noise, and a spectral region dominated by metabolic signals. See function "JBA_corDistribution()".
int	character vector indicating the method used to calculate the bin intensity. Possible values are: "sum", "mean", "max", "median".
cm	character vector specifying the correlation method ("pearson" or "spearman").
ef	numeric value establishing the maximum number of upfield or downfield variables that can be used to expand the seed. The maximum number of variables that can be added on each side of a given seed (size = st) is st*ef.
merge	character constant indicating whether highly correlated (correlation > mt) adjacent bins will be merged (i.e. integrated as a single bin).
mt	numeric value indicating the correlation threshold used to merge adjacent bins. This argument is ignored if merge = FALSE.

Details

JBA ("pJRES Binning Algorithm") is a new binning method designed to extend the applicability of SRV (Blaise et al., 2009) to pJRES data. The main steps of the JBA algorithm are described below:

1) The algorithm scans the NMR spectra (from low to high frequencies) and calculates the average correlation of st adjacent variables, using a sliding window of size one. This means that a given bin i starts at the NMR variable i and finishes at NMR variable with i + (st -1).

2) The vector of average correlations can be visualized as pseudo-NMR spectrum, displaying the average correlation values in the y-axis and the chemical shifts in the x-axis. This correlation-based spectrum is then scanned to identify local maxima passing the ct threshold. Each of these local maxima is used as seed that can be expanded by progressively aggregating upfield and downfield NMR variables, as long as the following criteria are met: (i) the average correlation of the bin remains equal or above ct; (ii) for a given upfield variable (vi), correlation (vi, vi+1) needs to be equal or higher than correlation (vi, vi-1); (iii) for a given downfield variable (vz), correlation (vz, vz+1) needs to be equal or lower than correlation (vz, vz-1).

3) The intensity of each bin is calculated as the mean, median, sum or maximum intensity of all variables within the bin. Notice that due to misalignments/signal overlap, it is possible that a single peak is split into several bins. These bins can be detected based on a given correlation threshold and integrated as a single bin.

Value

A list containing binned NMR data and information about the bins, as indicated below:

-The first element of the list ("all_clusters") reports the average correlation of st adjacent NMR variables along the chemical shift axis, using a sliding window of size one.

-The second element ("JBA_seeds") contains the local maxima (i.e. seeds) of the correlation-based spectrum along the chemical shift axis.

-The fourth element ("JBA_bins_expanded") indicates the bin edges after expanding the seeds by aggregating upfield and downfield NMR variables.

-The fourth element ("JBA_data") contains the binned NMR data.

References

Blaise, et al. (2009). Statistical recoupling prior to significance testing in nuclear magnetic resonance based metabonomics. Analytical Chemistry, 81, 6242-6251.

Examples

## Not available.

MWASTools documentation built on Nov. 8, 2020, 5:07 p.m.