Method for Calling Variants, False Positives

[arcolombo698]

Hello.

I have a question regarding the algorithm our lab uses for variant calling.
Essentially when calling the variants we call the variants for each chromosome individually. But this can cause errors (false positives) because if the reference map actually belongs to a different chromosome,say chromosome i, the algorithm will generate a score for the current chromosome having the variants called, say chromosome x.

However, for computational speed, splitting up the variant calling is more efficient because each node will do the process for each chromosome.

I am just looking for general advice.

The only way I can test this method is to finalize the VCF files, and use VCFtools to remove false positives.
-or I can use GATK, or Atlas2 to do a comparative analysis.

[yueluo]

Hi, just curious, did you
1) align your reads to individual chromosomes and then call variant, or 2) align reads to the whole genome and then call variant for individual chromosomes ?

[arcolombo698]

Calling Variants of aligned reads to the whole genome

Hello.

I aligned the reads to the entire genome, and then called the variants for each individual chromosome.

(#2)

[vivek_]

Its really not advisable to be splitting the mapping process by chromosomes. You will lose quite a bit of information and run into false positives in the latter stages, better idea to quicken this process is to split the reads into multiple files, align them to the reference and merge in the end.

Variant calling by chromosome is an entirely different issue and totally safe to divide this into multiple jobs by chromosome without worrying about false positives.

[arcolombo698]

I have a question the alignment process.

for a given sample, I normally perform quality control on the samples and afterward for a specific sample, I run the alignment on the paired trimmed samples (ie. running the alignments on the entire trimmed fastq files).

Are you suggesting that it is not ideal to run the alignment on an entire fastq file? but a partition of the fastq file? Does it matter if the project is DNA ? or RNAseq?

Thank you.
I am just confused because I normally run the alignments on a complete fastq file, and not a partition of a fastq file.

Is there a resource that you'd recommend?

[vivek_]

I have a question the alignment process.

for a given sample, I normally perform quality control on the samples and afterward for a specific sample, I run the alignment on the paired trimmed samples (ie. running the alignments on the entire trimmed fastq files).

Are you suggesting that it is not ideal to run the alignment on an entire fastq file? but a partition of the fastq file? Does it matter if the project is DNA ? or RNAseq?

Thank you.
I am just confused because I normally run the alignments on a complete fastq file, and not a partition of a fastq file.

Is there a resource that you'd recommend?

No, its not an issue if you decide to run the alignment of an entire fastq file against the entire reference sequence.

I was suggesting the preferred way of doing the task in parallel to may be utilize a HPC cluster if you have access to one. You'd rather split the fastq file than split the reference file for parallelism.

[arcolombo698]

thank you. makes sense.

is it more accurate to run an alignment on a smaller fastq file input into the aligner?