Floating point exception encountered

[Donovan Parks]

Hello,

I am finding MetaBAT to give extremely good bins on single metagenomes. Currently, I am using MetaBAT to bin a community comprised of 4 samples. I'm running MetaBAT through the shell script:

runMetaBat.sh scaffolds_500bp.fa.gz *.bam

This crashes with the following:

runMetaBat.sh: line 115: 42287 Floating point exception(core dumped) $MB $metabatopts --inFile $assembly --outFile $outname --abdFile ${depth}

I suspect the problem might be caused by some of my contigs having extremely low coverage and no reported variance. As such, I have played around with the --minCVSum, -x, and -m parameters to try and remove such contigs, but this doesn't resolve the issue.

I have placed the abundance and scaffold files on an FTP site at: https://data.ace.uq.edu.au/public/misc_downloads/

Regards, Donovan

[Don Kang]

Thanks Donovan for doing that. We'll take a look the issue.

[Don Kang]

Hi Donovan,

I've run your data with MetaBAT 0.25.3, but didn't reproduce your issue. Can you try with this command line as below to see if you still have the same issue?

ddkang@mc0537:/global/u1/d/ddkang/files/Donovan$ metabat -i metabat.scaffolds_500bp.fa -a metabat.coverage.txt -o bin -v
[00:00:00] Using p1 90%, p2 90%, p3 95%, minProb 80%, minBinned 40%, minCV 1.0, minContig 2500, minContigByCorr 1000, minCorr 92%, paired 0
[00:00:26] Finished reading 794909 contigs. Number of target contigs >= 2500 are 51650, and [1000 and 2500) are 254528
[00:00:28] Finished reading 306178 contigs (using 292722 including 241135 short contigs) and 4 coverages from metabat.coverage.txt
[00:00:49] Finished TNF calculation.
[00:02:16] Finished building a probabilistic graph. (292722 vertices and 13774806 edges)
[00:02:39] 1st round binning 100.0% (292722 of 292722)
[00:02:39] Leftover contigs before fish_more: 90.14% (263863 out of 292722)
[00:02:44] Leftover contigs after fish_more_by_friends_membership (roughly): 85.82% (251203 out of 292722), 111 bins
[00:02:44] Leftover contigs after fish_more (roughly): 85.78% (251109 out of 292722)
[00:02:44] Leftover contigs after fish_more_by_corr (roughly): 85.78% (251109 out of 292722)
[00:02:47] 24.07% (240027536 out of 997350751 bases) was binned.
Number of clusters formed: 111

[Donovan Parks]

Seems the problem is that my input scaffolds were compressed with gzip. The program "seems" to run fine with a compressed file as the resulting abundance file looks right, but the metabat executable is clearly unhappy with the compression.

[Don Kang]

Right. The compressed file was not expected for metabat. Maybe we should allow that later update. Thanks for reporting this!

[Don Kang]

Resolved.