What's left of 5 million reads

[Manu]

Hi,

I'm working with Illumina GAII data (5 million reads) from a 5' cap capture library. I did the mapping against a transcriptome with ELAND_standalone and now I'm left with 500000 reads that were mapped to a Gene-ID (after filtering). Most of the reads had more than one hit or had one or two mismatches. I used the default values for the filtering (not sure if I can change these with the standalone).

Now I just want to know if you get the same numbers (only 10 % mappable reads) with your data or if that number is too low.

[NextGenSeq]

10% is very low. BLAST or BLAT some of the unmapped reads to see if they are real or artifacts. If real obviously its a software processing error, if an artifact obviously a library problem.

[mattanswers]

Hi Manu,

I would try to download and use bowtie. It is much more flexible as to what bases you use to align as well as much faster. For me, it was a matter of hours for an alignment with ELAND as opposed to a couple of minutes using bowtie. So, with bowtie it is much easier to try different alignment parameters such as trimming 5' or 3' bases from your sequences. You can also print out a separate file of non-aligned sequences. It does not take too long to familiarize yourself with bowtie and in the long run seems well worth it.

Also, fastqc, http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc/ can be used with s_*_sequence.txt files and will quickly provide you with some quality assessment. It will tell you about the base qualities, GC content, and also provide an idea if there were adapter seqs or many duplicate reads as well as other stats.

[Manu]

I tried bowtie and bwa and both got better results. I think the ELAND filtering is very strict and I didn't find any information how to change the filtering parameter with the ELAND_standalone.