De novo trouble

[swapna]

Hello,

I have 1.3 million paired-end 454 reads for a bacterial genome with an expected size of 4.5 MB.

When I tried de novo assembly using the default Newbler parameters it estimated a genome size of 197 MB!! I got over 2000 contigs with an N50 of 2216 bases

Is there an input parameter for expected genome size? Or are there any other parameters that I should modify to get a better output?

Thanks,
Swapna

[maven]

Take it to your support rep. Newbler's genome size estimate are calculated directly from the coverage depth histogram immediately preceding it in the 454NewblerMetrics.txt file. So if that histogram is skewed for some reason the size estimate will be off. 2000 contigs for a 4.5 Mb genome is very bad, therefore you probably have something about your experiment that is causing it to contig poorly. The first thing that springs to mind is that your 1.3 million reads seems like way more than you really need. If they are 400 bp each and they all derive from your genome of interest, that's something like 115x. Newbler likes 20-35x the best, so try assembling less data. Second, if you have contaminants from other genomes then you could really be looking at a combination of contigs from your genome of interest, plus a whole lot of garbage from the contamination. The total contig length might give you a clue about this, as would the size distribution and coverage distribution of contigs (check 454ContigGraph.txt). Finally if the genome is highly repetitive or extremely biased in AT content (like P.falciparum, which is somethiing like 80% AT) then it might be really hard to get good contigging no matter what you do. Do any of these scenarios seem likely to you?