Only 2% mapped reads?????

[Jeannine]

Hey there,

I'm currently doing basic analysis for 3 different ChIP-seq experiments, i.e.mapping (using BWA) and peak calling (MACS). All 3 samples have been processed at the same time, give approx.38million reads each (131bp reads) and the sequence quality is the same (not extremely good but alright after trimming).
BUT, when I mapped the reads to the genome (b37), only 2% of the reads of one sample map, while I get approx.85% mapped reads with the other two samples. 2 % ???!!
How can that happen? Any suggestions what went wrong or where the problem is? Just by looking at the sequence reads in the fastq file I can't see any problems... Any help highly appreciated!!!!

[chadn737]

First thing I always do when a library seems to fail that badly is blast some of the reads to make sure they actually match the organism. I have identified many problems with libraries before using this approach.

[gringer]

I've had a yeast sample contaminated with human samples which fitted this pattern. I got back something like 1-2% of reads mapping to yeast mRNA, and eventually discovered (after a few days of frustration) that around 70% of the reads mapped to the human genome.

Now for mRNA reads I map to the SILVA rRNA database as something of a quality check, then give people stories about the probability of human rRNA contamination (e.g. if you accept that there was human contamination, then you also need to accept contamination from a lobster, a NZ weta, and a small bat).