Tabix

If tabix were a character, it would be the swordsman Roronoa Zoro from One Piece, slicing through your genome data to deliver you the result you want.

Tabix is part of Samtools. It indexes a tab-delimted genome position file (VCF file for example). Once indexed, it can quickly retrieve data from any part of the file without decompressing it. Thank you Heng Li 🙏 (tabix author). Here are some things you can do with tabix to a vcf.gz file.

First, you’ll want to index your VCF like this. This index acts lke a table of contents, helping tabix jump around the .gz file quickly.

The index file is called .tbi and will be put adjacent to your .gz file.

Now you can check for the existence of a variant like so.

Its important to remember, while you’re doing this, that a polymorphism being in the VCF file is not synonymous with being interesting or important or rare. There are lots of harmful / interesting / important polymorphisms in the reference genome. But more on that in another post.

You can look for variants in a region in the same way.

You can also do other useful / cool things with tabix, like return the header of a VCF.

tabix -H vcf.gz

Return chromosome names of a VCF.

tabix -l vcf.gz

You can also return variants found in regions listed in a file. The file can be a bed (.bed, .bed.gz, .bed.bgz) or a TAB-delimited file with CHROM, POS, and, optionally, POS_TO columns.

I like to use BED files. BED files have only 3 required columns: chromosome, start position, and end position. They have 9 more optional columns. The only optional column I use is the 4th one, which is name. I use it because it helps me remember what that locus is. More on the BED file format here. Here is a BED file I might use.

chr7    150999022    150999023   rs1799983
chr19   51354483     51354484    rs79338777
chr21   36146407     36146408    rs1056892

NOTE:This bed file is searching for three SNPs. Notice how the positions are one base apart? The position range in regions listed in a file need to start one base before where the expected variant is. That is different from when we search for a SNP directly like tabix test.vcf.gz chr7:150999023-150999023 for example, where the position range is from one position to the same position. If you were to do this in the BED file and a variant actually existed at position 150999023 on chromosome 7, it wouldn’t be returned. I don’t know why this is the case, but its certainly must-know behavior.

To return variants in regions listed in a file, enter:

tabix -R test.bed vcf.gz

Tabix doesn’t have a stdout option, so you would save the output to a file like this:

tabix -R test.bed vcf.gz > results.tsv

Careful with that .gz file …

gzip and bgzip are not the same thing. Wait … what is bgzip? Good question. bgzip is a compression algorithm that creates a series of gzip compressed blocks which are each 64kb in size. To go along with all these little blocks is whats called a “virtual offset”. The vitual offset is actually an unsigned (meaning nonnegative) 64 bit integer. And this integer holds information which acts like a map of the bgzip file. Its because of the “block + map” structure that bgzip files can be accessed without unzipping the entire file. And this is why tabix can access lets say chr1:452923-452923 without unzipping the whole variants.vcf.gz (a bgzipped file).

But why did I say be careful? Because bgzip files and gzip files both have the filename extension .gz. So how do you know if a .gz file is bgzipped or just gzipped? Because for the reasons explained above, you cannot access a part of a gzip file without unzipping it, which is why like tabix requires bgzip compression. Luckily its easy to tell. Just try to tabix it (create an index) and if its not bgzipped, tabix will tell you:

Learn more about bgzip and gzip in Peter Cock’s detailed post.