Hi Gongxin

Could you elaborate on what type of assemblies - transcript or genomic?

Dear Kristoffer,

Thanks.

I mean transcripts. And they were assembled by de novo RNA-seq assemblers.

For guided/de-novo assembled transcripts I would suggest you start by using a tool to predict whether the transcript is coding or not - and afterwards identify the ORF (either via one of the tools you mention of based on the longest ORF (>95% of proper annotated CDS are the longest)).

For predicting coding vs noncoding I usually use:

1) PhyloCSF which utilizes a comparative genomics apporach ( http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3117341/ )

2) CPAT which integrates multiple different metrics to predict the coding potential ( http://www.ncbi.nlm.nih.gov/pubmed/23335781 )

3) Additionally one could analyze the ORF for protein domains via f.x. pfam ( pfam.xfam.org ) and make a resonable argument that if a protein domain is found in the predicted ORF it is quite likely to be coding.

Hope this helps

Dear Kristoffer,

Thanks for your answers very much.

Your answers are clear. So they are helpful for me.