fitCalMaTeInternal: Algorithms to fit the CalMaTe model for a single SNP

In HenrikBengtsson/calmate: Improved Allele-Specific Copy Number of SNP Microarrays for Downstream Segmentation

Algorithms to fit the CalMaTe model for a single SNP

Description

Algorithms to fit the CalMaTe model for a single SNP. Note: These are internal functions of the package. They should not be used elsewhere.

Usage

  fitCalMaTeV1(dataT, references, fB1=1/3, fB2=2/3, maxIter=50, ...)
  fitCalMaTeV2(dataT, references, fB1=1/3, fB2=2/3, maxIter=50, ...)
  fitCalMaTeMedians(dataT, references, fB1=1/3, fB2=2/3,...)

Arguments

dataT	A 2xI numeric matrix of allele specific copy numbers (ASCNs), where 2 is the number alleles and I is the number of samples.
references	A integer vector with elements in [1,I] specifying which samples should be used as the reference set.
fB1, fB2	Thresholds for calling genotypes AA, AB, BB from the allele B fractions.
maxIter	The maximum number of iterations without converging before the algorithm quits.
...	Not used.

Value

Returns a 2xI numeric matrix of calibrated ASCNs.

Flavor v1

This is an early version (June 2010-January 2012) of the algorithm described in [1].

Flavor v2

This is the model and algorithm described in [1].

This version was introduced to decrease the number of "artificial outliers" introduced by CalMaTe for some SNPs due to non-converging or wreak-havoc estimates of the SNP effects. Flavor v2 differ from Flavor v1 as follows:

• The estimation of the model parameters are now done solely based on reference samples. In previous versions, some of the initial estimation steps were using also non-reference samples.

• For a small number of SNPs, the main CalMaTe scheme for estimating parameters would not converge or converge poorly. For such SNPs CalMaTe now falls back to using a plain median estimator, i.e. fitCalMaTeMedians().

• The above fallback estimator is also used in cases where all samples are identified to be homozygous.

The fitCalMaTeMedians() method is used as a fallback method by fitCalMaTeV2(). It fits CalMaTe without using the rlm function.

References

[1] M. Ortiz-Estevez, A. Aramburu, H. Bengtsson, P. Neuvial and A. Rubio, CalMaTe: A method and software to improve allele-specific copy number of SNP arrays for downstream segmentation, Bioinformatics, 2012 [PMC3381965].

See Also

These functions are called by calmateByThetaAB().

HenrikBengtsson/calmate documentation built on March 21, 2022, 6:32 p.m.