Vahid i gone through this also. it has limit of 1000 nucleotide but in genome sequence it will be more so it is not useful for identification of miRNA from genome sequence

Thanks for your suggestion..........

If you have not already done so check out this paper

http://bioinformatics.oxfordjournals.org/content/21/18/3610.full

Although this is not an answer to the original question posted, I have a related question. Since hundreds of mRNA targets are listed for every miRNA in mirase and most of them are not real targets to the specific miRNA, I want to know if there is an in silico approach to narrow down the targets to select a few relevant targets.

There are many tools to predict miRNA from genomic locations but they need small sequences they will not run on whole genome some of the tools are micropred, tripletSVM cshmm mireval etc but you cannot scan whole genome using these tools (mireval is a webserver).

To find relevant miRNA targets you can browse tarbase, ASRP (for arabidopsis miRNA targets). Also if you want to search the targets which is more relevant check the expression between miRNA and mRNA. you can do insilico by accessing GEO and by analyzing microarray data and NGS read data or best use PCR data to find the expression between miRNA and mRNA. In ideal case the correlation coefficient between miRNA and mRNA expression should be negative use all those targets which have correlation coefficient less than -0.5

what genome are you looking at? simple, moderate or complex i.e. genome size, repeat/transposon content and ploidy level? the strategy to identify miRNA de-novo should include these as important as well. Does your target genome has any close/near-close model relative also sequenced? Then you could also include comparative genomics based approaches.

MirPred

Zhang thank you for your help, but is there is software that runs n windows platform?