Extract mutational signatures from trinucletide context.

Share:

Description

Decompose a matrix of 96 substitution classes into n signatures.

Usage

1
2
extractSignatures(mat, n = NULL, nTry = 6, plotBestFitRes = FALSE,
  parallel = NULL)

Arguments

mat

Input matrix of diemnsion nx96 generated by trinucleotideMatrix

n

decompose matrix into n signatures. Default NULL. Tries to predict best value for n by running NMF on a range of values and chooses based on cophenetic correlation coefficient.

nTry

tries upto this number of signatures before choosing best n. Default 6.

plotBestFitRes

plots consensus heatmap for range of values tried. Default FALSE

parallel

calls to .opt argument of nmf. e.g, 'P4' for using 4 cores. See note on nmf for MAC users.

Details

This function decomposes a non-negative matrix into n signatures. Extracted signatures are compared against 21 experimentally validated signatures by calculating cosine similarity. See http://www.nature.com/nature/journal/v500/n7463/fig_tab/nature12477_F2.html for details. Please be noted that the original study described 21 validated signatures, however cosimc catalogue of cancer signatures has now reached ~30 signatures. Validated signatures lack some of, now well known signatures such Signature-22 (T>A strand bias occuring in liver), and this comparison might not include them. In that case you may have to manually infer the results.

Value

a list with decomposed scaled signatures, signature contributions in each sample and a cosine similarity table against validated signatures.

See Also

trinucleotideMatrix plotSignatures

Examples

1
2
3
4
5
6
## Not run: 
laml.tnm <- trinucleotideMatrix(maf = laml, ref_genome = 'hg19.fa', prefix = 'chr',
add = TRUE, useSyn = TRUE)
laml.sign <- extractSignatures(mat = laml.tnm, plotBestFitRes = FALSE)

## End(Not run)

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.