Clustering/mapping genes across de novo assemblies

[aprice67]

Hello,

I have two different strains of the same bacteria, strain A and strain B. I assembled both of these using trinity de novo, and I also have a reference genome for each strain, with reads aligned using bowtie. I want to find a way to make a mapping/clustering between strain A's de novo assembly, and strain B's de novo assembly. I also want to be able to make the same kind of mapping between strain A's de novo, and strain A's reference genome.

What I have:

• strain A: trinity de novo assemble (fasta)
• strain B: trinity de novo assemble (fasta)
• strain A: reads mapped to de novo assembly (bam)
• strain B: reads mapped to de novo assembly (bam)
• Strain A: reads mapped to reference genome via bowtie2 (bam)
• Strain B: reads mapped to reference genome via bowtie2 (bam)

What I want:

• Compare gene expression levels across assembly methods.
• Compare gene expression levels for same assembly methods, across different strains.

What I need:

• A method to identify which genes are which between these data!

So how can I get that mapping? I realize it wont be a 1-to-1 mapping, but with closely related sequences like this I could at least identify a majority of genes. If it's not something commonly done, can someone at least point me in a promising direction? Thanks very much for any suggestions you can offer.

[GenoMax]

Have you looked at Mauve for doing the genome-level comparisons? If you have annotation available for your reference and your strains are similar to the reference then that can make the comparisons easy. Are your assemblies in a single contig i.e. finished?

[aprice67]

Thanks for your reply.

I've been looking at Mauve now at your suggestion and it looks like it might be very useful if it does the kind of things I think it says it does. As for my annotations, all of the organisms I am using have published reference genomes available. As for my assemblies, I'm not exactly sure. I assembled a huge number of reads using trinity, I read that trinity is supposed to perform scaffolding, but I honestly hadn't been paying much attention to that. This is my first time doing any assembly work, so I'm going back now to check on those details. What I had been doing, was creating a bowtie index based on the trinity.fasta file, then mapping reads back with bowtie. Like this:

bowtie2-build Trinity.fasta Trinity

bowtie2 --all -x Trinity -1 forward_reads.fastq -2 reverse_reads.fastq -S aligned_reads.sam

I had assumed, incorrectly I expect, that I could do the same thing, just using the actual reference genome instead of the assembled data, and end up with two sam files that I could make comparisons on. Am I missing something huge here?



Anyhow, here is some information from one of my assemblies in case it helps. I'm going to continue looking into Mauve, I would welcome any suggestions or advice on what might be a good way to proceed and any pitfalls I might want to avoid. Thanks very much!


################################
## Counts of transcripts, etc.
################################
Total trinity 'genes': 2254
Total trinity transcripts: 2501
Percent GC: 46.47

########################################
Stats based on ALL transcript contigs:
########################################

Contig N10: 20428
Contig N20: 16367
Contig N30: 14168
Contig N40: 10881
Contig N50: 8121

Median contig length: 688
Average contig: 2780.90
Total assembled bases: 6955043


#####################################################
## Stats based on ONLY LONGEST ISOFORM per 'GENE':
#####################################################

Contig N10: 20428
Contig N20: 16483
Contig N30: 14250
Contig N40: 11095
Contig N50: 8404

Median contig length: 790
Average contig: 2950.57
Total assembled bases: 6650579

[GenoMax]

Trinity is not the right assembler for bacterial genomes. I suggest that you try SPAdes or Velvet. If you have a hugh amount of coverage you will be better off subsampling the data when you do assemblies. You should be able to generate a single contig (or a reasonably small number of contigs) easily.