Sequence Alignment Quality Control

[nilshomer]

As manual processing of sequencing data becomes impossible, it would be good to identify sequencing quality metrics that could/should be calculated/plotted within a sequencing pipeline.

What programs do you use to check the quality of your sequencing runs? Specifically, what metrics have you found useful pre-alignment, post-alignment, and post-variant-detection? What metrics/plots would you find useful if they were automatically generated?

[drio]

As manual processing of sequencing data becomes impossible, it would be good to identify sequencing quality metrics that could/should be calculated/plotted within a sequencing pipeline.

What programs do you use to check the quality of your sequencing runs? Specifically, what metrics have you found useful pre-alignment, post-alignment, and post-variant-detection? What metrics/plots would you find useful if they were automatically generated?

Here are some that come to my mind:

Pre-alignment:

1. Assuming we have available genotype data on the sample, some quick screening of the reads to find genotype concordance can be useful to avoid sample swaps.

Post-alignment:

1. Basic stats: % mapped uniquely mapped reads, % mapped reads, effective %mapped reads (after removing PCR duplicates), total throughput, effective throughput (after removing duplicates).

2. Error rate plot. For CS, before and after CS corrections.

3. For MP data: insert size distribution plot.

Post variant detection:

1. plot: ref/var coverage distribution.

Nils, which ones are you using on your end? Anyone?

[nilshomer]

Here are some that come to my mind:

Pre-alignment:

1. Assuming we have available genotype data on the sample, some quick screening of the reads to find genotype concordance can be useful to avoid sample swaps.

Post-alignment:

1. Basic stats: % mapped uniquely mapped reads, % mapped reads, effective %mapped reads (after removing PCR duplicates), total throughput, effective throughput (after removing duplicates).

2. Error rate plot. For CS, before and after CS corrections.

3. For MP data: insert size distribution plot.

Post variant detection:

1. plot: ref/var coverage distribution.

Nils, which ones are you using on your end? Anyone?

I have few those above implemented and more but I want to get a reasonable group of QC metrics before releasing. How would you compare genotypes before alignment? Would you look for specific sequences?

[Chipper]

Post alignment: compare enrichment / read distributions for different fractions of reads (high quality vs low quality, number of mismatches for reads in ChIP-seq peaks etc). Bin reads by average QV and compare % aligned reads for different aligners at different QV. For comparison between aligners I would also look at differences in coverage over various repeats, what % reads are uniquely placed by only one aligner and how many reads are placed at different positions.

[nilshomer]

Post alignment: compare enrichment / read distributions for different fractions of reads (high quality vs low quality, number of mismatches for reads in ChIP-seq peaks etc). Bin reads by average QV and compare % aligned reads for different aligners at different QV. For comparison between aligners I would also look at differences in coverage over various repeats, what % reads are uniquely placed by only one aligner and how many reads are placed at different positions.

I do not mean to compare aligners, but evaluate the alignment itself.