About Mapping Efficiency of Bisulfite aligners

Question

[chaudharyc61]

Hello Everyone I have mapped my own data from Arabidopsis meiocytes cells, with BS-seeker2 and bismark with both of modes (End-toEnd and local). The mapping efficiency is obviously different as BS-Seeker2 uses local alignment.

My question is :- Is it okay to use local alignment for BS-seq data with seed of approx 20 bps and mismatches of only 0.01 while aligning? heres the table of mapping efficiency

1. BS-seeker2 66.13% (local)
2. bismark 38% (end-to-end which is default) 66% ( local )

Points to remember : My data is generated from meiocytes. local alignment can make alignment of reads to different locations by random chance, which can affect the downstream analysis. And my data is Non-directional.

Thanks Chandan kumar

Answer 1

[Devon Ryan]

Yes, it is completely OK to use local alignment with BS-seq data. I'm a bit surprised that it took bismark so long to finally support it, since bs-seeker2 and bwa-meth have long used local alignment (and generally perform better).

Make sure to set a decent MAPQ threshold when extracting methylation metrics, as this will handle the multimapper issue.

[chaudharyc61]

Thank you so much for clearing my doubt.

What do you mean by setting MAPQ while extracting methylation ???

[ATpoint]

Take only reads with a certain mapping quality (MAPQ), for example 20, as the mapping quality is an inverse probability that a read maps to multiple location equally well. The higher MAPQ, the more likely it maps "uniquely". Vice versa, the lower MAPQ, the more likely it is that it maps to multiple locations equally well. This then would make the alignment less reliable, which is not what you want.

[chaudharyc61]

Hello Devon Ryan

I have another issue related to the same problem. As i am using bismark for mapping and the result format is in bowie2 format, The best possible alignment score in end-to-end mode is 0, and The best possible score in local mode equals the match bonus times the length of the read. And I have done the mapping in both of the modes and the MAPQ (fifth column) is not according to the bowtie2 format What i mean is in local mode the highest score is 4 only but the number of uniquely aligned reads are thrice of time compared to end-to-end mode.

I am very much confused which mode should i use to carry on.?

[Devon Ryan]

Are you sure the highest MAPQ is 4 and not 40? I generally prefer local alignment for data like this.

[chaudharyc61]

yes because i tried extracting reads with higher MAPQ values using

samtools view -b -q 5 Aligned_reads.sam -o Quality5.bam

And i got zero reads in the output file.

[Devon Ryan]

That sounds like a bug in bismark, make sure you're using the most recent version.