Analyze Copy Number Signatures with sigminer

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Exploring copy number signatures with recently developed approach have been described at The repertoire of copy number alteration signatures in human cancer.

A more general introduction please read Extract, Analyze and Visualize Mutational Signatures with Sigminer.

library(sigminer)

For this analysis, data with six columns are required.

Generate allele-specific copy number profile

load(system.file("extdata", "toy_segTab.RData",
  package = "sigminer", mustWork = TRUE
))

set.seed(1234)
segTabs$minor_cn <- sample(c(0, 1), size = nrow(segTabs), replace = TRUE)
cn <- read_copynumber(segTabs,
  seg_cols = c("chromosome", "start", "end", "segVal"),
  genome_measure = "wg", complement = TRUE, add_loh = TRUE
)
cn
cn@data

Classify the segments with Steele et al method

If you want to try other type of copy number signatures, change the method argument.

tally_s <- sig_tally(cn, method = "S")

str(tally_s$all_matrices, max.level = 1)

Find de novo signatures

sig_denovo = sig_auto_extract(tally_s$all_matrices$CN_48)
head(sig_denovo$Signature)

Refit (19) reference signatures

This directly calculates the contribution of 19 reference signatures.

act_refit = sig_fit(t(tally_s$all_matrices$CN_48), sig_index = "ALL", sig_db = "CNS_TCGA")

We can use some threshold to keep really contributed signautres.

act_refit2 = act_refit[apply(act_refit, 1, function(x) sum(x) > 0.1),]

rownames(act_refit2)

Plot signatures

For de novo signatures:

show_sig_profile(sig_denovo, mode = "copynumber", method = "S", style = "cosmic")

Show the activity/exposure.

show_sig_exposure(sig_denovo)

For reference signatures, you can just select what you want:

show_sig_profile(
  get_sig_db("CNS_TCGA")$db[, rownames(act_refit2)],
  style = "cosmic", 
  mode = "copynumber", method = "S", check_sig_names = FALSE)

Similarly for showing activity.

show_sig_exposure(act_refit2)

NOTE that this case shows relatively large difference with different approaches, so you need to pick based on your data size/quality and double-check the results. In general, for small-size data set, the refitting approach is recommended.

Signature assignment

To assign the de-novo signatures to reference signatures, we use cosine similarity.

get_sig_similarity(sig_denovo, sig_db = "CNS_TCGA")


Try the sigminer package in your browser

Any scripts or data that you put into this service are public.

sigminer documentation built on May 29, 2024, 3:11 a.m.