Redundans not producing expected output when trying to reduce metagenomic assemblies

[Jesse McNichol]

Hi Leszek,

I've been trying to use redundans to reduce and quantify the redundancy in metagenomic assemblies I'm working with. I've run into a few issues:

-contigs.reduced.fa.hetero.tsv is always empty
-contigs.reduced.fa.hist.png is always empty

Even if I don't set the --noreduction flag, it prints out the following config information with the -v flag:

Options: Namespace(fasta='180124_all_concatenated_cd-hit_culled.fasta', fastq=[], identity=0.51, iters=2, joins=5, limit=0.2, linkratio=0.7, log=<open file '<stderr>', mode 'w' at 0x7f5d772171e0>, longreads=[], mapq=10, minLength=200, nocleaning=True, nogapclosing=True, norearrangements=False, noreduction=True, noscaffolding=True, outdir='180202_test_cd-hit_culled_redundans_default', overlap=0.8, reference='', resume=False, threads=4, verbose=True)

Even though this flag is reported as above, I do get good reduction of about 95% for shorter contigs (<1000bp) in some samples (which is what I expected intuitively). For others, it does almost nothing (maybe it removes a few contigs but removes <1% for sure). This might be a real result but I'm just not sure whether I'm causing the unpredictable behaviour by using redundans for a case where the complexity is much greater than what you designed the software for. I expect there is more allelic variation compared to a single genome assembly since we're sequencing complex and mixed natural communities. Basically, I'm just not sure whether I'm running up into some limit of the software for making multiple comparisons/alignments.

If you have any insights, I'd be grateful to hear it. Can provide logs/input files if you'd like.

Best,

Jesse

[l.p.p...@gmail.com]

Hi Jesse, 

Honestly, I've never tried Redundans on microbiome data, but it can be an interesting application! 

You may observe removal of short contigs even without reduction, as contigs shorter than 200bp (minLength=200) are skipped, as those complicate scaffolding and usually resulting gaps can be later easily closed with PE/MP reads. 

First of all, which version/commit are you running (git log | head -n5)? 

I've been trying to use redundans to reduce and quantify the redundancy in metagenomic assemblies I'm working with. I've run into a few issues:

-contigs.reduced.fa.hetero.tsv is always empty
-contigs.reduced.fa.hist.png is always empty

Even if I don't set the --noreduction flag, it prints out the following config information with the -v flag:

Options: Namespace(fasta='180124_all_concatenated_cd-hit_culled.fasta', fastq=[], identity=0.51, iters=2, joins=5, limit=0.2, linkratio=0.7, log=<open file '<stderr>', mode 'w' at 0x7f5d772171e0>, longreads=[], mapq=10, minLength=200, nocleaning=True, nogapclosing=True, norearrangements=False, noreduction=True, noscaffolding=True, outdir='180202_test_cd-hit_culled_redundans_default', overlap=0.8, reference='', resume=False, threads=4, verbose=True)

That's correct, when you set --noreduction this parameter becomes False and reduction is not performed (I know it's counterintuitive:/). 

 

Even though this flag is reported as above, I do get good reduction of about 95% for shorter contigs (<1000bp) in some samples (which is what I expected intuitively). For others, it does almost nothing (maybe it removes a few contigs but removes <1% for sure). This might be a real result but I'm just not sure whether I'm causing the unpredictable behaviour by using redundans for a case where the complexity is much greater than what you designed the software for. I expect there is more allelic variation compared to a single genome assembly since we're sequencing complex and mixed natural communities. Basically, I'm just not sure whether I'm running up into some limit of the software for making multiple comparisons/alignments.

If you have any insights, I'd be grateful to hear it. Can provide logs/input files if you'd like.

It may be related to data itself. 

Which de novo assembler are you using?

Could you send me contigs.reduced.fa.hist.png and logs/statistics for good and bad cases? 

Bests, 

L. 

Best,

Jesse

--
You received this message because you are subscribed to the Google Groups "Redundans" group.
To unsubscribe from this group and stop receiving emails from it, send an email to redundans+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/redundans/fc91fafb-9750-4416-9f58-0b93c2bd0c95%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.