High number of N in reads

[papori]

Hi all!
is there any consensus on th way to filter out reads that have high percents of "N"?
For example:
if more than 50% of the characters are "N", then you filter out the read.

And without any connection to the type of application that you work on.

if there is, there is a different between illumina platforms?

Thanks!
Pap

[GenoMax]

Sequencing has improved to a point where there should be very few reads with long stretches of N's (unless there was some issue with base calling). Certainly there is no reason to have 50% N's in a single read.

There will generally be a small number of reads with N's at the beginning of reads (specially in first tile).

[papori]

you are right,
but there are still many reasons to have "N".. not just the sequencing platform but the sample quality..
in fact, there still many experiments in the SRA that are from GAII/hiseq200 with many N in the reads..
i am asking this because i am facing with that at the moment..

Any idea when the read is helpless?

[GenoMax]

This is not a good answer but it may depend on what you are trying to do. You may be able to get away with N's if you are aligning to a reference and have sufficient depth of coverage. Not so much with any de novo assembly.

I would be weary of using reads that have more than 5-8% N's. If the N's are at the ends of the reads you could trim them but if if they are in the middle of the read ....