Creating Hg19 Reference Index

Question

[Davy]

I am creating the index reference in order to align my fastq reads. One question I have is should I include all the "funny" chromosomes like the chr6hapcox or chrUnxxxxxxx?

I will be using GATK downstream for other parts of the analysis, and I know it doesn't like when the chromosomes are out of order, so what should I do as regards these non-canonical chromosomes which don't have (at least to me) an implicit ordering?

Answer 1

[Pierre Lindenbaum]

See Heng Li 's page : http://lh3lh3.users.sourceforge.net/humanref.shtml

For variant discovery, RNA-seq and ChIP-seq, it is recommended to use the entire primary assembly, including assembled chromosomes AND unlocalized/unplaced contigs, for the purpose of read mapping. Not including unlocalized and unplaced contigs potentially leads to more mapping errors.

[Davy]

I realise this, but the problem is it will cause GATK to throw an error when parsing the BAM files. Any ideas on how to get GATK to not break?

[JC]

you can map with your favorite tool and filter the funny hits from the SAM/BAM file after that.

Answer 2

[Rok]

If you have your chromosome in separate fasta files you should merge everything into one fasta file (whole genome). Using Create Sequence Dictionary from Picard Tools you create a dictionary for this whole genome. This is going to store order of chromosomes in the whole genome file.

When you do mapping with a whole genome fasta file as an index most of the mapping software should produce SAM/BAM header that is the same as the dictionary file. If it happens to be different (TopHat sometimes seems to sort chromosomes a bit differently) you can use ReorderSam from Picard tools to sort SAM/BAM file in the same fashion as it's in the dictionary, be careful you provide ReorderSam with correct dictionary file.