VCF calling for individual chromosomes, domain reduction

[arcolombo698]

Hello.

I am just seeing if this method for VCF creating for variant calls is correct.
Right now I keep the fastq as an entire file, and then align the whole fastq file to individual chromosomes, and this produces a VCF file for each chromosome that I later merge together.

However this can cause errors if the alignment has a poor map to the individual chromosome because it will generate a forced alignment, even if it maps poorly to the chromosome. (ie the VCF call will show a score to the poor map of the fastq to the chromosome if there is only a single chromosome the fastq is being aligned to:: (ie. reducing the domain of the fastq to a single chromsome at a time)


Is it better not to reduce the alignment map to a single chromosome at a time and instead map to the whole chromosome at once?

[Jeremy]

Yes, it is much better to align to the entire genome in a single step. If you want to reduce the memory load then separate the fastq into a few smaller files and align them separately instead of separating the genome like that.

[arcolombo698]

I am not sure how to do this.

how do you split the fastq files? how to align the whole genome?

when I align the fastq using Tophat, I align both reads at the same time.

[Jeremy]

A fastq is just a text file with one read every 4 lines, if you have paired end data it then two text files. The reads should be in the same order in both files by default so just split the files into smaller files making sure that the number of lines is a factor of 4. Actually, splitting probably isn't necessary anyway.

[arcolombo698]

Thank you
I am familiar with the alignment practice, and am interested in the variant calls.
How do I create the VCF files, without doing the variant calling for a single chromosome at a time?

[Jeremy]

Not much information to go on there. But the first step is to map against the whole genome not single chromosomes, so if you are using tophat then make a bowtie reference of the whole genome, map to that then use the resulting bam to call variants using whatever program you use which will call variants for the whole genome.

[arcolombo698]

Vcf

yes. this is the protocol I follow.

I do have an HG19 human reference genome that I align the fastq files to.
after I use same tools to index the accepted hits.bam file and use mpileup -g command to call the variants.

what parameters do you use to call the variants? do you use mpileup -g ?