raw illumina reads or assembled contigs for Taxonomy

[Girija]

Hello!

I am having whole genome sequenced metagenomic data that is assembled using meta-velevt. I want to check taxonomic classification using MgRAST. I need to know that whether I should use raw reads or assembled contigs for taxonomy classification as I am interested in showing functional annotation using assembled contigs.

Please help me in clearing this confusion that which is better way for taxonomy classification; raw reads or assembled contigs?

Best regards

[Brian Bushnell]

Assembled contigs data should be much better than raw reads, if available.

[ankit4035]

It is entirely up to you as both the methods are technically correct...

I will list out the merits and demerits of both and then you can decide it for yourself..

1. Accuracy
   Assembling the metagenomic reads is quite problematic. While, various tools are available specific for metagenomic data its efficiency depends on the nature of your sample. More complex metagenome(having lots of strain level diversity) is more difficult to assemble and often generates chimeric contigs.
   
   
2. Abundance
   Raw reads provide a means to relatively quantify your taxa or functions as compared to assembled data. Since, all the reads merge together to form contigs, actual abundance of annotated taxa(or function) cannot be determined.
   
   
3. False annotation
   Short reads are not favorable for annotation and can often lead to false positives. Also, some of the times ambiguity remains for accurate annotation (However, it can be eliminated by approaches like least common ancestor(LCA) but it also decreases the sensitivity of your data). In such scenarios, having a longer contigs are favorable.

On a personal note, I am using quality filtered reads for annotation using MG-RAST.

Hope it helps...