Output filtering

[Roman Hillje]

Hey everybody!

Just wanted to quickly ask to understand if, by default, all output filtering rules described in the manual are applied?

Example: 

I want to align 50bp reads from an RNAseq experiment to the reference genome. Among many others, there are these two rules for output filtering:

--outFilterMismatchNmax [int, default "10"]

    alignment will be output only if it has no more mismatches than this value

--outFilterMismatchNoverLmax [float, default "0.3"]

    alignment will be output only if its ratio of mismatches to mapped length is less than or equal to this value

Does that mean that in theory a total of 10 mismatches would be allowed per read, but since the reads are only 50 bp long the second rule will reduce that limit to a maximum of 1 tolerated mismatch (50 * 0.03 = 1.5) per read?

I hope I understood that correctly.

Thanks,

Roman

[Alexander Dobin]

Hi Roman,

all default and user-defined rules are applied with the "AND" logic, i.e. the most stringent rule wins.

However, in your example, the value is 0.3, so for 50b reads it will limit MMs at 15, which means that the other rule (<=10MM) is more stringent and it wins.

Cheers

Alex

[Roman Hillje]

Excellent, thank you!