SFF: PE or not PE? That is my question.

[dan]

Is there a simple way to determine if a given SFF file contains data for Paired End (PE) or Single end (SE) reads?

I guess I could try matching the standard linker sequence to some of the reads, but it would be nice if the information were directly readable from the file. I tried looking at the manifest using sffinfo, but their was nothing there that suggested PE or not PE.

I found that if I run the SFF through gsMapper, it will create ACCNO_left and ACCNO_right version of the reads, representing the left and right ends of a given PE read. However, its a pain to run all my SFF's through gsMapper just to determine which ones it (somehow) determines to be PE.

If I ran SSAHA2 to find the linker, for example, what would be the command line?

Is their a better tool to choose?

Sorry for the Noobs...

Cheers,

[dan]

I just found a dirty hack to check if my SFF file is PE or not PE (titanium):

grep -c "TCGTATAACTTCGTATAATGTATGCTATACGAAGTTATTACG" *.sff.fna


or:

for file in *.sff; do
echo $file
sffinfo -s $file | grep -c ""TCGTATAACTTCGTATAATGTATGCTATACGAAGTTATTACG
done


This 'hack' gave a few 10s of thousands of 'hits' in PE files (depending on the size of the region) vs. a reassuring 0 hits in non PE files (30 files tested, 0 instances so far).

[dan]

xxxxxxxxxx

[maubp]

Is there a simple way to determine if a given SFF file contains data for Paired End (PE) or Single end (SE) reads?

I guess I could try matching the standard linker sequence to some of the reads, but it would be nice if the information were directly readable from the file. I tried looking at the manifest using sffinfo, but their was nothing there that suggested PE or not PE.

I have been through the SFF format in detail, and the only place this information could be recorded is in the Roche XML manifest, but thus far it isn't. I believe this is because for the sequencing machine a single-end and a paired-end run are handled in exactly the same way - the only difference is in the sample preparation.

So yes, searching for the appropriate paired end linker sequence (different for old 454 and the newer Titanium kits) is your only option. I know that sff_extract calls SSAHA2 to detect the linkers. However, a simple string search for a perfect match will be good enough (even though this will miss a lot of almost perfect linkers) to decide if a Roche SFF file is paired end data or not.

[flxlex]

Another fast way is to start an assembly using newbler (e.g. on the command line) using your sff file(s) of interest, and wait for the output mentioning something like this:

Indexing XXXXXXXX.sff...
-> XXXXXX reads, XXXXXXXX bases, XXXXXX paired reads.

Files with only shotgun reads will not have the 'paired reads' number. This only works, however, if at least 25% of the reads in the sff file contain the paired end linker (as per the manual).