deconvolve

Share:

Description

deconvolve is a linear deconvolution procedure for STR DNA mixtures.

Usage

1
2
3
deconvolve(nC, mixData, refData = NULL, condOrder = NULL,
  locsel_Mix = NULL, locsel_Ref = NULL, eps = 100, zeroMx = FALSE,
  threshT = 50, verbose = FALSE)

Arguments

nC

Number of contributors in model.

mixData

Evidence object with list elements adata[[i]] and hdata[[i]]. Each element has a loci-list with list-element 'i' storing qualitative data in 'adata' and quantitative data in 'hdata'.

refData

Reference objects with list element [[i]][[s]] where list-element 's' is the reference index and the list-element 'i is the loci index where the qualitative data is stored as a length two vector.

condOrder

Specify conditioning references from refData (must be consistent order). For instance condOrder=(0,2,1,0) means that we restrict the model such that Ref2 and Ref3 are respectively conditioned as 2. contributor and 1. contributor in the model.

locsel_Mix

Boolean-vector with Selected loci in mixData to deconvolve. locsel_Mix=NULL; accepts all loci.

locsel_Ref

Boolean-matrix for specifying conditional loci (row) for each reference (column).locsel_Ref=NULL; accepts all loci.

eps

Number of best combinations to keep during the search.

zeroMx

boolean of allowing zero mixture proportion as an estimate for any contributors.

threshT

Imputet quantitative value when conditioned reference alleles are non-observed.

verbose

Boolean for whether in-process information should be printed

Details

The procedure optimizes the mixture proportion simultaneous with combined genotypes by assuming the STR response variation as normal distributed. The criterion for optimization is the error distance Mahalanobis Distance (MD) between the fitting model and observed responses.

Conditioning on referenced genotypes is possible. Selection of conditioned loci for each of the references may be specified. Unobserved alleles from references will be imputed as observed alleles with the input threshold as the quantitative information. Non-selected or empty observed loci will return NA as genotype combinations and not treated in model.

The search strategy is called keepElite which optimizes over all loci simultaniously by storing the 'eps' best fitted combinations during the search. The function also returns the optimized marginal result (each loci optimized).

The covariance structures taking all loci into account assumes a compound symmetry structure which takes the number of alleles and peak heights into account (this ensures 'proportion of variance').

The user may choose whether combinations giving zero mixture propotion (gives overfitting model) for any contributors are accepted.

Value

Optimized deconvolution model object.

simpleList

Table of loci independent optimizations. Uses independent covariance structure.

pList

Resultlist of optimized combinations, mixture proportions and error-distances (MD).

locinames

Name of loci in mixData

result1

Tabled optimized results in format 1.

result2

Tabled optimized results in format 2.

data

All data used as input in analysis.

options

Input parameters used in analysis.

Author(s)

Oyvind Bleka <Oyvind.Bleka.at.fhi.no>

References

Tvedebrink,T, et.al.(2012). Identifying contributors of DNA mixtures by means of quantitative information of STR typing. Journal of Computational Biology, 19(7),887-902.