How to assign gene names to your sequences during gene annotation?

Hi guys, I've finished annotating my de novo assembled transcriptome with blast2go and exported the results to a fasta file. The format of the fasta file is like this:

>Seq1|e3 ubiquitin-protein ligase rnf213 isoform x2|GO:1234 GO:4567 GO:7890

SEQUENCE

>Seq2|transcription factor hivep2|GO:4321 GO:7654 GO:8970

SEQUENCE

The gene names/description and GO terms (separated by "|") are assigned by blast2go. My situation is that I am not allowed to use the gene names assigned by blast2go. How do you guys get gene names for you sequences? What methods/tools do you use to assign accurate gene names to you sequences?

annotation trancriptome blast2go

• 234 views • link

updated 10.3 years ago by Michael Dondrup

• written 10.3 years ago by ngs101.com

Curious about what you mean by I am not allowed to use the gene names assigned by blast2go. IP reasons or you think they are not correct?

• 1 views • link

written 10.3 years ago by GenoMax

Haha, not me but my advisor doesn't trust blast2go. I am fine with blast2go.

• 1 views • link

written 10.3 years ago by ngs101.com

sed 's#Seq#ThisIsNotBlast2GO_I_promise#'

• 1 views • link

written 10.3 years ago by cpcantalapiedra

What is this? I don't get it.

• 1 views • link

written 10.3 years ago by ngs101.com

A post in jest. sed is a streaming editor that will allow you to change every occurrence of a word (here Seq) with something else (the rest of the big term) in a file.

• 1 views • link

written 10.3 years ago by GenoMax

OK, it's just a joke?

• 1 views • link

written 10.3 years ago by ngs101.com

1 answer

Your supervisor has a point here. Using a gene symbol without any addition means that you are quite sure about their function. A pure inference by sequence similarity is often not considered sufficient. On the other hand, most functional annotation is assigned based on computational methods. If you look at human genes for example (e.g. https://www.ncbi.nlm.nih.gov/gene/9961 ) you will find a lot of information about the long standing curation and annotation process, that is based on various evidence. If you want to annotate a homolog inferred by Blast2Go you need to be a bit more reluctant in your first attempt to name it, you could name it "Putative ... gene", or "Similar to ... gene [homo sapiens]".

• 1 views • link

written 10.3 years ago by Michael Dondrup

Thank you for you reply, Michael. Blast2go is not based on"A pure inference by sequence similarity". I've spent so much time on the algorithm of B2G.

For GO terms assignment, it uses an annotation rule which integrates seq similarity, evidence code and hierarchy structure of GO terms to compute the best GO terms for a sequence.

For gene name/description assignment, it uses an algorithm called Blast Description Annotator which computes the most frequent names among the blast hits.

Unlike blast2go, other tools put so much weight on the sequence similarity when annotating sequences. The annotation rule and Blast Description Annotator algorithms are quite reasonable to me. Unfortunately, my advisor doesn't trust them. So I am wondering if there are any other ways to get gene names instead of using the Blast Description Annotator algorithm.

• 1 views • link

written 10.3 years ago by ngs101.com

What is the final aim of your experiment? Is annotation of the assembly the end point or are you going to use this for differential expression analysis or some other purpose?

• 1 views • link

written 10.3 years ago by GenoMax

I am trying to make a reference transcriptome of my species and then use RNAseq data to do some differential expression analysis based on this annotated transcriptome.

• 1 views • link

written 10.3 years ago by ngs101.com

Perhaps you should focus on doing the DE analysis first (not having a gene name right won't prevent you from doing that). Once you have a small number of genes of interest you could come back and refine/reaffirm B2G annotation just for those. Confirming annotation on all genes could become a very lengthy/involved process.

• 1 views • link

written 10.3 years ago by GenoMax

Constructing a reference transcriptome is one of my projects. I need to get it done to graduate. Besides, my other projects are dependent on this reference transcriptome. What my advisor wants is to assign accurate gene names to each of my sequences without using B2G. What is the common way of doing it without the help of B2G?

• 1 views • link

written 10.3 years ago by ngs101.com

B2G is encapsulating the process one would normally follow for annotating an unknown sequence in a pipeline.

Briefly .. Doing multiple blast (blastn, blastx) searches. Pulling out sequences that match. Building MSA to identify regions of homology/motifs/domains .. so on and so forth.
As @Michael said at the beginning of this discussion, the only way you would conclusively prove that a sequence is really "gene x" would require doing additional experiments (e.g. a knock-out).

• 1 views • link

written 10.3 years ago by GenoMax

I think the point I wanted to make is that a lot of lab biologists won't fully trust the computational prediction, however refined, and ofc blast2go uses an advanced algorithm to weight evidence. InterPro is another important "meta search engine", and there are many more, and of course you can use the GO term assignments for your analysis. When you said, my supervisor doesn't allow me to use gene names or gene symbols, then this is more about the correct use of scientific terminology in the field of molecular biology and genome annotation, and I think you should discuss with your supervisor what the acceptable lingo would be, and what would be an acceptable level of evidence to switch from "conserved protein, similar to shikimate kinase" to "shikimate kinase".

Certainly, your DE analysis can be done without assigning gene names in the first place, and GO terms can be used if your methods section mentions that GO terms were predicted by Blast2GO.

• 1 views • link

written 10.3 years ago by Michael Dondrup

Thank you, Michael and genomax2. I guess there is no easy way to assign accurate gene names to your sequences without wet lab evidence. I may not make myself clear. It is not my advisor doesn't trust computational prediction. It is he doesn't trust the algorithm B2G used to assign gene names to the sequences, i.e. the Blast Description Annotator algorithm. My advisor's point is he wants to come up with a method that is better than blast2go to predict gene names.

What he has done is he spent about 8 months blasting 3000 sequences manually on NCBI website and assigned a gene name to each of these 3000 sequences based on his blast results. He then compared these gene names to the names assigned by B2G. Any difference he found in gene names assigned by these 2 methods is deemed as inaccurate prediction by B2G because he believes his manual blast is highly accurate. It is this part that confused me.

• 1 views • link

written 10.3 years ago by ngs101.com