Comparing assemblies using BLAST or mapper?

[morning latte]

Greetings,

I want to compare metagenomes generated by Illumina HiSeq. Metagenomic sequences were assembled so I want to compare different assemblies. I have heard that BLAST could be used for comparing assemblies. In this case, what blast program can be suggested, blastn or blastx? Can mappers be used (bow tie, BWA, etc)? Any suggestion or advice would be appreciated!

[sphil]

what do you mean by "compare"? I doubt that mappers like BWA.. are capable of doing this. They are desinged to work with short sequences, <120bp, so they won't compare sequences of, approx. 1kb, very well. Blast is one shot you can also try someting like mauve, assuming that you say that the two assemblies are one genome. Mauve will show you a 'full genome' alignment.

[GenoMax]

If you are trying to "compare" assemblies of significant similarities then MUAVE may be an option to try: http://gel.ahabs.wisc.edu/mauve/

LASTZ can also be used if you want to do pair-wise alignments: http://www.bx.psu.edu/~rsharris/lastz/newer/

[Brian Bushnell]

I like to use quast for comparing assemblies. It has a special version, metaquast, that is designed for metagenomes.

I also routinely evaluate assemblies using BBMap.

bbmap.sh ref=assembly1.fa in=assembly2.fa nodisk out=mapped.sam

This will shred assembly2 into 500bp fragments, map them to assembly1, and generate statistics about how much mapped (you can make the fragment length shorter with the 'fastareadlen=X' flag). If you do it both ways you will see how much of the assembly of each is contained by the other.

Also, mapping the reads to each assembly will give you nice statistics about the percentage of of reads map, the percent that matched the reference perfectly, the percent that were ambiguous, and the error rates (insertions, deletions, substitutions), which allows you to infer the quality of the assembly. The best metagenomic assembly will typically allow the highest mapping percent with the lowest error rates, lowest ambiguity rates, and smallest size.

The included assembly stats program will also give you an overview of the assembly continuity:

stats.sh assembly.fa

[morning latte]

Dear all,

I really appreciate all the suggestions and advices. This is really helpful. I will try some of them now. Thank you again.

[morning latte]

Dear Brian Bushnell,

So I have tried with metaquast according to one of your suggestions even though I ran into some problems (contigs_reports directory contains nothing) and trying to figure out. I have several environmental metagenomes and each of them was assembled. In its publication, it does not describe metaquast command as it came out recently I guess. Just to make sure, metaquast performs like aligner for metagenomes? Most of contig size in my metagenomes are short and the average contig size is about 550 bp. Could you give me a bit more detail about this metaquast command if you have any experience? I appreciate your advice in advance.

[Brian Bushnell]

Unfortunately, I can't really help you there - I've only used quast, not metaquast. And yes, it will align assemblies to each other (as well as doing other things, like gene counting).

[GenoMax]

@Morning latte: At least one person from the group that wrote Quast is on SeqAnswers. If you create a separate thread with your question about Metaquast it should generate a response.

[morning latte]

Thank you Brian Bushnell and GenoMax!