Filter out supplementary aligned reads for variant calling or not

Question

[gb]

Dear all,

I am using BWA-mem and freebayes to do some variant calling and I was wondering if I should remove the supplementary aligned reads or not. I can not find much info about this.

[ATpoint]

In general it is good practice to not do any custom filtering unless the manual of the tool you use specifically recommends it. You should have duplicates marked in the BAM. Any proper variant caller will then internally choose the reads that based on its underlying model or strategy are suitable. The README (https://github.com/ekg/freebayes/blob/master/README.md) does not mention any prefiltering so you should not do it.

[gb]

I think that is a good point. For now I will not worry about this.

[ATpoint]

If you later during analysis of the results see that something is wrong you can still dig deeper if supp. alns might be a factor but I doubt so. I learned over the years that one should not overthink things and only start to debug if there is evidence for something going wrong. It might cost a lot of time without making a positive difference.