FreekManders
commented
3 years ago This normally can't be turned off. Here is a version of the reading functions that perform this filtering step, with the filter removed. If you load these functions in R, it should probably work. (I can imagine a function like plot_rainfall might have some issues though.) Can you confirm if this works?
read_vcfs_as_granges <- function(vcf_files,
sample_names,
genome,
group = c("auto+sex", "auto", "sex", "circular", "all", "none"),
type = c("snv", "indel", "dbs", "mbs", "all"),
change_seqnames = TRUE,
predefined_dbs_mbs = FALSE) {
# Match arguments
type <- match.arg(type)
group <- match.arg(group)
# Check sample names
if (length(vcf_files) != length(sample_names)) {
stop("Please provide the same number of sample names as VCF files", call. = FALSE)
}
# Get the reference genome
tryCatch(
error = function(cnd) {
stop("Please provide the name of a BSgenome object.", call. = FALSE)
},
{
genome <- BSgenome::getBSgenome(genome)
}
)
# Read vcfs
grl <- purrr::map(vcf_files, .read_single_vcf_as_grange, genome, group, change_seqnames) %>%
GenomicRanges::GRangesList()
# Filter for mutation type
if (type != "all") {
grl <- get_mut_type(grl, type, predefined_dbs_mbs)
}
# Set the provided names for the samples.
names(grl) <- sample_names
return(grl)
}
.read_single_vcf_as_grange <- function(vcf_file, genome, group, change_seqnames) {
# Use VariantAnnotation's readVcf, but only store the
# GRanges information in memory. This speeds up the
# loading significantly.
# Muffle the warning about duplicate keys.
withCallingHandlers(
{
gr <- GenomicRanges::granges(VariantAnnotation::readVcf(vcf_file))
},
warning = function(w) {
if (grepl("duplicate keys in header will be forced to unique rownames", conditionMessage(w))) {
invokeRestart("muffleWarning")
}
}
)
# Throw a warning when a file is empty.
# Return a empty GR, to prevent errors with changing the seqlevels.
if (!length(gr)) {
warning(paste0(
"There were 0 variants (before filtering) found in the vcf file: ", vcf_file,
"\nYou might want to remove this sample from your analysis."
), call. = FALSE)
return(gr)
}
# Convert to a single chromosome naming standard.
if (change_seqnames == TRUE) {
tryCatch(
error = function(cnd) {
message(conditionMessage(cnd))
stop("The seqlevelStyle of the vcf could not be changed to that of the reference.
You can run this function with `change_seqnames = F` and `group = 'none'`,
to prevent this error.
However, you then have to make sure that the seqnames (chromosome names) are
the same between your vcfs and the reference BSgenome object.
(The message of the internal error causing this problem is shown above.)",
call. = FALSE
)
},
{
GenomeInfoDb::seqlevelsStyle(gr) <- GenomeInfoDb::seqlevelsStyle(genome)[1]
}
)
}
# Change the genome name of the granges
genome_name <- GenomeInfoDb::genome(genome)[[1]]
GenomeInfoDb::genome(gr) <- genome_name
# Filter for variants with the correct seqlevels
if (group != "none") {
tryCatch(
error = function(cnd) {
message(conditionMessage(cnd))
stop("The vcf could not be filtered for the specific seqlevels group.
You can run this function with `group = 'none'`, to prevent this error.
(The message of the internal error causing this problem is shown above.)",
call. = FALSE
)
},
{
gr <- .filter_seqlevels(gr, group, genome)
}
)
}
return(gr)
}
.filter_seqlevels <- function(gr, group, genome) {
groups <- c()
# These variables are needed to extract the possible seqlevels
ref_style <- GenomeInfoDb::seqlevelsStyle(genome)
ref_organism <- GenomeInfoDb::organism(genome)
if (group == "auto+sex") {
groups <- c(
GenomeInfoDb::extractSeqlevelsByGroup(
species = ref_organism,
style = ref_style,
group = "auto"
),
GenomeInfoDb::extractSeqlevelsByGroup(
species = ref_organism,
style = ref_style,
group = "sex"
)
)
# In some cases, the seqlevelsStyle returns multiple styles.
# In this case, we need to do a little more work to extract
# a vector of seqlevels from it.
groups_names <- names(groups)
if (!is.null(groups_names)) {
# The seqlevels in the groups are now duplicated.
# The following code deduplicates the list items, so that
# creating a data frame will work as expected.
unique_names <- unique(groups_names)
groups <- lapply(unique_names, function(x) groups[groups_names == x])
groups <- lapply(groups, unlist, recursive = FALSE)
# In case there are multiple styles applied, we only use the first.
groups <- unique(as.vector(groups[[1]]))
}
}
else {
groups <- GenomeInfoDb::extractSeqlevelsByGroup(
species = ref_organism,
style = ref_style,
group = group
)
groups <- unique(as.vector(t(groups)))
}
# The provided VCF files may not contain all chromosomes that are
# available in the reference genome. Therefore, we only take the
# chromosomes that are actually available in the VCF file,
# belonging to the filter group.
groups <- BiocGenerics::intersect(groups, GenomeInfoDb::seqlevels(gr))
# We use 'pruning.mode = "tidy"' to minimize the deleterious effect
# on variants, yet, remove all variants that aren't in the filter
# group. By default, keepSeqlevels would produce an error.
gr <- GenomeInfoDb::keepSeqlevels(gr, groups, pruning.mode = "tidy")
return(gr)
}
mcdoz2010
gevro
Hello there, I would like to use Mutational Patterns on Illumina data from a DNA virus genome amplified from clinical samples, in order to identify the mutational processes that generate new virus variants in vivo. The sequencing depth is about 10 000x and I use Lofreq to generate the vcf files. In this case, there can be two mutations at the same site, but when the vcf files are loaded into a GRangesList, I get the following error message;
There were 2 duplicated variants in vcf file:.vcf They have been filtered out.
Is there any way I can disable this filtering step?
Thanks for a great package by the way!
Dorian