Is bcf just a bgzipped vcf?

Question

[endrebak]

99% sure, but could not find the answer on google. Creating this q for other bcf newbs.

[venu]

• VCF AND BCF

Answer 1

[Paul]

BCF is binary counterpart of VCF file. Please look here for description. Similar to SAM vs BAM.

[Pierre Lindenbaum]

and, sadly, BCF is not always bgzipped... (raw bcf)

There is also “raw BCF” which bcftools outputs with -Ou. Historically HTSJDK has only supported raw BCF not real BCF https://t.co/EqDIAxtsWH

— John Marshall (@jomarnz) September 15, 2017

[jena]

"Historically"? The linked github issue is still open. And e.g. bcftools still internally converts all formats to uncompressed BCF for processing.

Can somebody please explain to me why should I use BCF? I only read that it should be faster to parse, without further explanation. The same source also says I should use VCF for data sharing and processing by custom scripts (DOI: 10.1093/gigascience/giab008).

Why is BCF faster to parse? Why not just use indexed & (b)gzipped VCF? That would have benefits of both being small, fast and portable. What am I missing?

[Pierre Lindenbaum]

Can somebody please explain to me why should I use BCF? I only read that it should be faster to parse, without further explanation.

vcf is a text file that can be compressed (vcf.gz).

BCF is not a text format but a BINARY format which is de-facto faster to read.

e.g: how to read an integer in C / text (yeah I could use fscanf too )

int n=0,c;
while((c=fgetc(in))!=EOF && isdigit(c)) { n= n*10 + (c-'0')}

in C/binary

fread(in,sizeof(int),1,&n);