I have always faced a problem while analyzing small RNAseq data, at the step of adapter trimming.
I use fastx_clipper to clip adapters. The problem is that you can't be really sure of very small alignments between adapter and reads because they may not really be originating from the adapters, which means that you should specify a lower limit of alignment for clipping. I usually set it as 5 (intuitively).
But if a small piece of sequence really came from the adapter then it will remain and there is no way to clip it without any doubt.
The real problem comes during aligning the reads to the reference sequence. Aligners like bowtie (which i prefer to use), generally have user defined argument for number of allowed mismatches, which generally shouldn't exceed 4 both for performance and stringency.
Subsequently, you might lose a really valuable read.
An example: A certain unknown miRNA (annotated as one in mirbase) is 28 nt long. If I clip adapters with minimum alignment of 5 then I end up having 2-3 nucleotides of adapter in one of the highly expressed miRNA in that tissue, which then goes undetected because of mismatch in alignment.
Alternative
To avoid this problem I sometimes trim the reads to around 25nt (for miRNA profiling). This creates a new problem: You can't really distinguish whether the read came from a pre-miRNA (a longer RNA) or from mature-miRNA (smaller RNA that arises by processing of pre-miRNA)
If I already have some idea of what I am going to get after analysis, I can tweak it somehow (for e.g by grep-ing the first 2/3/4 letters of the adapter sequence in the miRNA database to find out the chance of random occurrence of adapter like sequences). But if I am looking for something new it wont help.
Does anyone have an experience or an idea about how to solve this problem?
One question: How long are the unclipped reads you use? The ones I used till now where something ~36nt. But with that length you would expect an alignment of ~10nt and not 5, or less.
Well yes..you are right.. i didnt think about it.. my mistake..
but isnt illumina small RNAseq adapter : TCGTATGCCGTCTTCTGCTTGT.... I sometimes end up getting reads which have some extra nucleotides in the end which are neither from pre-miRNA nor from adapter. Some don't even align back to the genome.
For example this read "AGGCAGTGTAGTTAGCTGATTGCAT" (post clipping) doesn't align to genome. Removing last 2 bases aligns it to mmu-mir-34c-5p. "AT" is not coming from the adapter. If i remove the last three bases this is the top miRNA. If i dont then it drops to 20th. The phred scores in that region are decent.
Perhaps these nucleotides you want to cut are actually correct?!? MicroRNA 3′ nontemplated nucleotide additions are actually quite common and can be found in all short RNAseq experiments. I would recommend you to allow more errors for the mapping and analyse what you get, instead of changing your data (and nature) by hand. :) (read e.g. http://rnajournal.cshlp.org/content/17/10/1795.full)
thats what i finally did.. i use this algorithm called mirdeep.. i just modified it a little so that it runs bowtie in the n-alignment mode.. considering the biology of miRNAs i allow no mismatches in first 10-15 nt but allow 2-3 mismatches (corresponding to phred score 80-120) in the remaining region.. is this strategy okay ??