Dear Wannes,

Without claiming to be an expert, I think that there are usually a number of reasons for this observation. First of all, de novo transcriptome assemblies tend to be rather fragmented, frequently resulting in contigs which are actually only part of the transcript, i.e. fully or mostly consist of UTR regions. These contigs will of course not have any Blast hit. In part, the percentage of "no Blast hits" thus also depends on the quality of the assembly. Furthermore, depending on the e-Value cutoff you use, transcripts coding for rather short and not highly conserved peptides can be below this e-Value cutoff. Especially for some groups of invertebrates, the number of closely related species with well-annotated reference genomes and thus the occurrence of hits with high similarity can be low. Last but not least, with more recent RNAseq approaches and deep sequencing, you will also capture transcripts which do not encode for proteins, some of which stem from genomic "background transcription". And (although their definition is a bit unclear without comparison to a sufficient number of closely related species) let´s not forget the so-called "orphan genes", which can be species-specific, with overall low homology to known genes.

Well, I am pretty certain I have not completely covered all of the possible explanations for this observation, but I hope this might help a bit.

Best regards,

Heiko

Heiko mentioned many important points.

May I ask how many reads you generated, what the size of the reads is what the base calling quality is, what the size of the target genome is, what the sequence technology used was and whether there was any size selection, etc.

There are many more points that would influence the assembly quality and in turn the mapping.

Hi Jens and Heiko

Sorry for the late reply!

Thank you for your valuable answers/comments. They were very useful to me.

Regards

Wannes