De Novo assembly of a plant transcriptome

[raonyguimaraes]

Hello all,

I work with Bioinformatics in a lab and our group received the task of mounting a plant transcriptome using sequences from 454.

We still don't have the genome sequenced so we are now defining our strategies for this project.

The first problem is related with the assembly of the sequences.

We used Newbler and Mira for the assembly and now we are lacking of metrics to compare this 2 softwares in order to decide which one is better and why.

After the assembly I think the next obvious thing to do is to align this transcriptome with the genome of arabdopsis and may be use this to decide which assembly was better and why.

Another question would be about which aligner we should use for this.

Does anyone have a suggestion or any experience to share about this project?

Thanks for help.

[circlez]

Regarding to the comparison of assemblers, here is a paper that might help.
Comparing de novo assemblers for 454 transcriptome data

http://www.ncbi.nlm.nih.gov/pubmed/20950480

I used Newbler 2.5 to assembly my data, which looks quite good.

[raonyguimaraes]

Thanks a lot !

I also found this website http://www.plantagora.org/ and a lot of metrics I can use to evaluate my assemblies

Check it out:

http://www.plantagora.org/metrics/

http://www.plantagora.org/Plantagora_Engine/assembly_evaluation.html

The following metrics were gathered as part of this project: total number of contigs, contig N50, total contig length, average contig length, largest contig length, contigs > 1kb, contigs > 5kb, number of scaffolds, total scaffold length, average scaffold length, largest scaffold length, and scaffold N50.

looks promising...

[MG1655]

Hi raon,
I've been doing de novo plant transcriptome as well (although with illumina reads) and have had much luck with trinityrnaseq. http://trinityrnaseq.sourceforge.net/

Also, plantgdb has a lot of resources (such as a download for all of the plant protein sequences known)

http://www.plantgdb.org/download/download.php/

[Jenzo]

MG1655, may I ask if your data was from a normalized or non-normalized cDNA bank?
Thanks!

[MG1655]

MG1655, may I ask if your data was from a normalized or non-normalized cDNA bank?
Thanks!

Hi Jenzo, our cDNA was non-normalized because we also wanted to look at differential expression.

[jmartin127]

Aligning your de novo assembled transcriptome to a genomic reference can help to determine which assembly is better. Presumably, whatever you are sequencing is a closely related plant to arabidopsis?

Regardless, Blat (http://genome.ucsc.edu/cgi-bin/hgBlat?command=start) is a good tool for splice-aligning assembled transcripts to a genome.

Hello all,

After the assembly I think the next obvious thing to do is to align this transcriptome with the genome of arabdopsis and may be use this to decide which assembly was better and why.

Another question would be about which aligner we should use for this.

[dilaraally]

de novo assembled transcriptome to a genomic reference

I am also doing a de novo assembly of a plant transcriptome. I used ABySS and CLC Bio. One thing I did after assembling the transcriptome is to do a legacy blast search. Not all contigs will necessarily be plant - could be fungi, bacteria, and virus. If you only blast to Arabidopsis you might pick up a highly conserved gene but whose top hit was actually not a plant but a bacteria- blasting only to Arabidopsis doesn't allow you to select out contamination.

My pipeline is the following:

blastn to entire ncbi database (locally)
|
remove non-plant contigs - keep ones that had top 10 hits to plants and ones that had no hit
|
blastx
|
with the final set of contigs, annotate using closest hit in blast search.