I've recently completed RNA-seq of 5 pilot human samples using the Takara/ClonTech SMARTer Stranded Total RNA-seq kit v2 - Pico input mammalian and Illumina HiSeq 4000 and have had troubling results. My samples are of low input quantity and so I had to chose a kit optimised for small input amounts.
The main problem is that the mapping was very poor for 3 of the samples, and on examination of the FastQC report the striking result for these samples was that the per base sequence content had a strange repeated 'saw-tooth' pattern (see figure 1).
The prominence of this pattern was inversely proportional to the mapping % i.e. the worse the mapping, the more prominent the 'saw-tooth' pattern was. The effect was not related to quality or quantity of RNA input. This seems to imply that there was insertion of trinucleotide repeat elements that are contaminating the reads. SMART technology relies on the terminal transferase activity of the kit's specific reverse transcriptase, which adds 3 nucleotides to the cDNA strand when it breaks off the template RNA strand. These 3 nucleotides are targeted by a second primer that causes reverse transcriptase to switch templates from the RNA strand to the primer and reverse transcribe the primer. This leads to a single stranded cDNA with primers on both ends that can be targeted by PCR primers for amplification. I wonder whether the kit could have introduced cytosine- (and guanine-) rich repeats somehow...?
Has anyone experienced this problem? If so, is there anything that can prevent it?
I'd be very grateful for any help!