Hello, I am looking for a calculation of gene coverage from the BAM file or any other file used in whole-exome data analysis. Please tell me if anyone knows about this. Thanks in advance.
Hello, I am looking for a calculation of gene coverage from the BAM file or any other file used in whole-exome data analysis. Please tell me if anyone knows about this. Thanks in advance.
One way, via bam2bed
and bedmap
:
$ bam2bed --reduced < reads.bam | bedmap --echo --count --delim '\t' genes.bed - > answer.bed
The count of reads overlapping a gene by one or more bases will be in the last column of answer.bed
.
To generate genes.bed
, this will depend on your assembly and how you name chromosomes. Gencode is one source for human and mouse gene annotations in GFF format, which can be converted to BED via gff2bed
, e.g. for hg38
:
$ wget -qO- ftp://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_38/gencode.v38.annotation.gff3.gz \
| gunzip --stdout - \
| awk '$3 == "gene"' - \
| gff2bed - \
> genes.bed
use a tool like featureCounts:
featureCounts: an efficient general purpose program for assigning sequence reads to genomic features
If you want more statistics than just counts of reads that overlap gene annotations, perhaps the following answer will help.
Starting with a generic BAM file, it can be converted to BED via bam2bed
.
The 13th column (QUAL) can be converted from its printable characters back to an average Phred quality score, and then swapped with the fifth (score) column, to following the UCSC specification for BED files.
This numerical value can subsequently be used for mapping, applying score statistical functions with bedmap
(shown further below):
$ bam2bed < reads.bam | awk -v FS="\t" -v OFS="\t" 'BEGIN{ for (n=0; n<256; n++) ord[sprintf("%c",n)]=n }{ old=$5; l=split($13,q,""); s=0; for(n=1; n<=l; n++) { s+=ord[q[n]]-33; } $5=s/l; $13=old; print $0; }' > reads.bed
You can start with genes as described in another answer in this thread, which are in a file called genes.bed
.
Now you can map your BED file containing genes-of-interest to the reads from the BAM file:
$ bedmap --skip-unmapped --delim '\t' --echo --count --bases-uniq --echo-ref-size --bases-uniq-f --mean genes.bed reads.bed > answer.bed
The file answer.bed
will contain summary statistics for reads mapped to genes and look something like this, unique to the makeup of your genes and reads:
$ head answer.bed
chr1 199064 200374 ... 404 1310 1310 1.000000 25
chr1 6853080 6870040 ... 204 9044 16960 0.533255 20
chr1 8004008 8027135 ... 507 22308 23127 0.964587 20
chr1 8061658 8098833 ... 366 37175 37175 1.000000 25
chr1 8174215 8211962 ... 369 34253 37747 0.907436 35
chr1 8873384 8890703 ... 434 17319 17319 1.000000 50
chr1 8874708 8890922 ... 420 16214 16214 1.000000 45
chr1 15833039 15850691 ... 271 17652 17652 1.000000 50
chr1 16643023 16645543 ... 266 2520 2520 1.000000 45
chr1 16665212 16667822 ... 281 2610 2610 1.000000 30
The first three columns are the intervals defined by genes. These and metadata columns come from using --echo
.
Columns in between will be metadata about the gene (symbol, strand, etc.) and depend on the source of gene annotations.
The last five columns come from the remaining options (--count
, --bases-uniq
, etc.) and are as follows:
The order of the specified bedmap
operands --count
, --bases-uniq
, --echo-ref-size
, --bases-uniq-f
, --mean
write the aforementioned five column values in that same order.
Other operations are available and can be added to the bedmap
command, if useful. Additionally, if one base of read coverage is too lenient, additional operands are available to further constrain the overlap criterion.
Run bedmap --help
and take a look at the "Overlap Options" and other sections, or take a look at the online documentation for a more complete walkthrough.
Have you checked some of the previous related discussions?
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
I got this answer:
Please tell me if 412 is the gene coverage.
412 reads overlap that element.
So this is variant supporting reads, not gene coverage, right?
what is your definition of "gene coverage"?
coverage typically means multiple features (say reads) covering a gene, the count is how many of the features covered the interval
In case you do not only want to count reads overlapping genes, I added a second answer that shows how to calculate additional statistics.
If you do not want coverage of reads over genes, please edit your question so that others may be able to help.