MetagenomicDeconvolution
Auth: Rogan Carr
Date: July 2013
About
-----
This software deconvolves Sample Taxa Abundances and Sample Element Abundances to predict the most likely distribution of the elements among the taxa, using the least squares regression and non-negative least squares regression implementations that are built in the base install of Matlab.
This software implements fully functional version of the Metagenomic Deconvolution Framework (MetaDecon) proposed by Rogan Carr, Shai Shen-Orr, and Elhanan Borenstein. However, this distribution is meant to serve as a basis for customization. For example, different "Masking" choices or regression solutions can be easily implemented.
Inputs
------
TaxaAbundaceFile - nSamples x nTaxa - tab delimited file containing taxa abundances for each sample
ElementAbundaceFile - nSamples x nElements - tab delimited file containing element abundances for each sample
ElementFile - 1 x nElements - tab delimeted file containing all element names
constantElementName - the name of the constant element
constantElementLength - the (approximate) length (or copy number) of the constant element in each taxa
outPrefix - the file prefix for all output
(optional) 'MaskAbundance', 'number between 0 and 1'
only deconvolve elements with a mean abundance at least the product of this number and the abundance of the most abundant element in any samples.
example: 'MaskAbundance',0.001 -> only deconvolve elements with a mean abundance at least 1/1000th that of the most abundant element in any sample.
Outputs
-------
outPrefix.LeastSquares.dat - the genome reconstruction for each taxa using least squares
outPrefix.NNLeastSquares.dat - the genome reconstruction for each taxa using non-negative least squares
outPrefix.Mask.dat - 1 x nElements - a list of 1s or 0s describing whether the element was solved for or not, respectively