Why there is some discrepancy between blastx and blastp outputs?

Question

[seta]

Hi all,

I have done a de novo transcriptome assembly for a non-model plant. I have then tried to predict ORFs using Transdecoder tool. Finally, I removed asterisk characters in the file1.transdecoder.pep by the command of sed -i 's/*//g' file.fasta and did blastp against some databases. But, I found that there is much differences between the blastp and blastx (that I also conducted it for the same databases) outputs. In fact, the maximum of ortholog hit ratio (OHR) for blastx was 1 while it was 0.5 for blastp. As you know, the OHR near to 1 represent the full-transcript and highly desirable. Could you please share your opinion about it and let me know which is wrong for blastp?

Thanks so much for you help.

Answer 1

[Michael]

There isn't anthing wrong realy with the blastp approach. But this case is a good example why it is often better to use blastx. ORF predictions can be unreliable on assembled fragmented transcripts. The contigs can e.g. contain frameshifts, and 6-frame translations cannot compensate for those. If the stop codon is not included in the contig things get worse, the ORF predictor might also pick the wrong reading frame.

[seta]

Thanks for your response. So it refer to ORF prediction; however the translated contig sequences is required for searching conserved domain, for example. Among several ORF prediction tool, which software do you recommend for this tasks?