Low mapping for genome and transcriptome

[neethav]

Hi all,
I have a set of 75bp SE illumina reads.My problem is that for some samples I am able to map almost 85% of the reads,but for some the mapping percentage is very low,about 20-30% (using MAQ,bwa,bowtie).I tried using bowtie options to trim,increase the mismatches etc.,with no significant increase in mapping.I also ran fastqc on mapped and unmapped reads but it looks very similar.I blasted some of the unmapped sequences,and some seems to be rRNA genes.Is there a way to find out what exactly is different between these libraries? Can I include such samples with high mapping and low mapping as biological replicates (or as test Vs control) for differential gene expression?
Thanks!

[mnkyboy]

Try making a ribosomal index and see what percentage of your samples are ribosomal. There may have been an inconsistency in the library prep.

Also what are the samples? For example in some of our viral systems we see a very large subset of the sample can be viral RNA.

[neethav]

My samples are from aphids(insect).I don't understand why the reads from rRNA are not mapping to the transcriptome (it has rRNA genes in it).

[kopi-o]

Could it be that the rRNA genes are repeated or redundant, so that the reads map to several different loci and are filtered out for that reason? Just speculating here.

[chadn737]

When you blast these reads is the entire read rRNA or only part of it? My first suspicion when I see samples that map at such a low rate with high quality sequence is adapter sequence in the read.

[kopi-o]

My first suspicion was adapter sequence contamination too ... but the OP wrote that they had run FastQC and seen similar results for the good and the bad samples (or perhaps I misunderstood). If it was adapter contamination, there should be a noticeable difference in the overrepresented sequences from FastQC.

[pbluescript]

Could you post some details of your library prep? I had a sample that ended up being contaminated with RNA from other sources. When I ran trinity and blasted the contigs, I ended up seeing a lot of stuff from other species. Mapping the reads to the contigs showed that >40% of the sample likely came from contamination.