regularizeMatrix: Careful regularization (pseudocount addition) to a position...

In profileScoreDist: Profile score distributions

Description

Carries out the regularization suggested by Rahmann et al. This lets each column in the regularized matrix be a linear combination of the column in the non-regularized matrix and rho, the overall base distribution of all positions. The weighting of the linear combination is determined by the parameter E in a non-trivial way, see Rahmann et al. for more information. A default value E=1.5 usually works well.

Usage

regularizeMatrix(motif, E = 1.5)

Arguments

motif	A position count matrix; each column a position and each row a base corresponding to A, C, G, T. This order is assumed, unless the rows are correspondingly named in a different order.
E	Weighting parameter between 0 and 3 for the regularization.

Value

The regularized matrix

References

Rahmann, S., Mueller, T., and Vingron, M. (2003). On the power of profiles for transcription factor binding site detection. Stat Appl Genet Mol Biol 2, Article7.

Examples

data(INR)
regularizeMatrix(INR)

Example output

       [,1]         [,2]         [,3]      [,4]         [,5]      [,6]
A  49.60074   0.02837893 288.03350163  26.45590  77.16267735  89.37062
C  48.82428 303.03893853   0.04596735  81.62553  95.22320845 148.69457
G  69.35925   0.01697079   0.02003420 116.27263   0.09728214  59.37778
T 137.63467   0.02998173  15.03539375  80.48164 131.17186511  95.63407
       [,7]      [,8]
A  46.55883  53.27450
C  87.13886 100.49292
G  73.93219  57.95817
T 101.64687 104.45944

profileScoreDist documentation built on Nov. 8, 2020, 5:49 p.m.