Transcript mapping stats from Cuffcompare - Does this look right to you?

[avi]

Hi all,

I am working on some RNAseq data (Single end reads,36 bp from an Illumina instrument) from a prostrate cancer cell line. All I have for this is a Fasta file of all the reads.

I have assembled the reads using Tophat and Cufflinks, and then ran Cuffcompare to look at the quality of transcriptome reconstruction. This was the profile of transfrags I got.

Category     No.of transfrags    % of total
Match	          1533	               1.73
Novel	          3561	               4.02
Contained	  24080	               27.18
Repeat	           0	                0
Intronic	  10115	               11.42
Polymerase        1889	               2.13
 run-on	
Intergenic	  28752	               32.46
Overlap on        14340	               16.19
opp.strand	
Total	          88580	               100

[I just grep'ed the tmap file to find no of rows with each class code]

I am new to RNAseq data, so I have no idea what to expect. But I find it surprising to see that only 1.73% of the total transfrags matched to a known transcript. And that over 32% mapped to intergenic regions. Even accounting for the fact that it is a cancer cell line and some amount of changes are to be expected.

I was hoping someone with experience could take a look at this and give their opinion. Are these kind of numbers common..? Or does this mean the data I got has some problems?

Also, in general.. are there any standard quality assurance steps I can use to check RNAseq data?

Would greatly appreciate any help that I can get on this..

thanks..!