miRNA differential expression software

[yueli]

Hi,

I'm currently analyzing small-RNA-seq data coming from 2 different samples which only have *.fasta format. Unfortunately they do no have replicates.

Which software that I can use to analyze their differential expression?

Following is the format of my data. t0000001 is ID, and 537124 is the reads number.

Any suggestions are appreciate!

li



>t0000001_x537124
CCCGACCTCAGATCAGATGA
>t0000002_x340107
TCACCGGGTAGACATTCATTAT
>t0000003_x268409
TGAAAGACATGGGTAGTGAGAT
>t0000004_x172657
CGATATGTGGTAATTTGGATGA
>t0000005_x154782
AGAGGTAGTGATTCAAAAAGTT
>t0000006_x140076
TCACCGGGTAGACATTCATTATA
>t0000007_x131590
TGAGATCACTATGAAAGCTGG
>t0000008_x125368
TGAGGTAGAATGTTGGATGACT
>t0000009_x120455
TGAGTATTGCATCAAGAACCGA
>t0000010_x95804
TCCCTGAGACCATTGACTGCAT
>t0000011_x78210
TGGACGGAAGTGTAATGAGGGT
>t0000012_x73438
TGAGGTAGATTGTTGGATGACT
>t0000013_x68884
CCCGACCTCAGATCAGATG
>t0000014_x60358
TGAAAGACACAGGTAGTGGGACA
>t0000015_x59786
TGGAATGTCGAGAAATATGCAT
>t0000016_x44935
TCCCTGAGACCATTGACT

[diego diaz]

Usually it's not recommended analyze differential expression when you don't have replicates, but if you want to try anyway, you could map your reads to your reference genome with bowtie (bowtie supports fasta format), then count the number of reads per microRNA gene with htseq-count and finally perform differential expression analysis with Deseq2.

In this thread, Deseq2's author (Simon Anders) explains how to perform the differential expression analysis when you don't have replicates!

http://seqanswers.com/forums/showthread.php?t=31036


To map your reads with bowtie you could use the following configuration,

-m 1 and -v 0, with m 1, all reads with more than one valid alignment are suppressed (it is a way to keep only uniquely mapped reads) and v is used to allow 0 mismatch.

Alternatively, you could use the following configuration

-l 15 , -n 0, -k 1 --best, -l and -n specify that n mismatch are allowed in the first l bases, and k specify the number of alignment reported. --best is to report the best, if more than one valid alignment exists


I hope it will be useful!

[yueli]

Hi,

Diego diaz.

Thanks a lot! I try to use Deseq2 to perform my analysis,

Li