miRNA-seq HiSeq2000 expression profiling problem

[mirandaseq]

I have a problem doing gene expression profiling of microRNAs. I have data of a miRNA-seq experiment on a HiSeq 2000 from the same rat tissue at 5 different time points. After adaptor trimming and aligning, the expression level of a given miRNA (from control, i.e. non-experimental tissue) is fluctuating across the different time points whereas one would expect the same expression level for all time points.
Attached is a plot showing the gene expression of one miRNA on a time course scale.
The following command was used to trim the adaptors:

./fastx_clipper -i /raw.fastq/miRNA.fastq -o /trimmed.adapter/miRNA.trimmed.adapter.fastq -a TGGAATTCTCGGGTGCCAAGGAACTCCAGTCAC -l 15 -M 20 -c

This command was used for alignment in bowtie:

bowtie -k 2 -v 0 -S -p 16 -q --solexa1.3-quals /ref/ref.genome /trimmed.adapter/miRNA.trimmed.adapter.fastq /align/miRNA.trimmed.adapter.sam

We tried to align the trimmed reads against known miRNAs (from mirBase) and against the genome and had nearly the same results. We also filtered out reads that are from other non-coding RNA classes (such as rRNA, tRNA, snRNA etc). Also, discarding reads shorter than 15 bp and longer than 30 bp yielded the same results, that is variable expression level for a given miRNA in control animals.

Any help is appreciated!

[volks]

how do you calculate the RPKMs ?

[mirandaseq]

read count/(total read count * miRNA length) * 1E6

[NicoBxl]

So what is the question ?

[volks]

read count/(total read count * miRNA length) * 1E6

total read count means reads mapped to miRNAs?

[mirandaseq]

Yes, this is what it means