Rcorrector dropping reads?

[macmanes]

I've uncovered a strange behavior that I'm having trouble understanding.. RCorrector installed via conda. I end up having lost a few R2 reads.

rcorrector                1.0.2                         1    bioconda

The command

perl run_rcorrector.pl -1 test.1.fq -2 test.2.fq

Output: Note the right number of reads are processed

Put the kmers into bloom filter
jellyfish bc -m 23 -s 100000000 -C -t 1 -o tmp_8946be6f4ed713dd539613da8d3abdf8.bc test.1.fq test.2.fq
Count the kmers in the bloom filter
jellyfish count -m 23 -s 100000 -C -t 1 --bc tmp_8946be6f4ed713dd539613da8d3abdf8.bc -o tmp_8946be6f4ed713dd539613da8d3abdf8.mer_counts test.1.fq test.2.fq
Dump the kmers
jellyfish dump -L 2 tmp_8946be6f4ed713dd539613da8d3abdf8.mer_counts > tmp_8946be6f4ed713dd539613da8d3abdf8.jf_dump
Error correction
/pylon5/mc3bg6p/macmanes/conda/Oyster_River_Protocol/software/anaconda/install/envs/orp_v2/bin/rcorrector   -p test.1.fq test.2.fq -c tmp_8946be6f4ed713dd539613da8d3abdf8.jf_dump
Stored 81770 kmers
Weak kmer threshold rate: 0.004975 (estimated from 0.950/1 of the chosen kmers)
Bad quality threshold is :
Processed 53898 reads
    Corrected 18524 bases.

Counting corrected reads: Note a few R2 reads lost

grep -c @DRR test.1.cor.fq
26949

grep -c @DRR test.2.cor.fq
26265

Counting original reads: 53898 total reads.

grep -c @DRR test.1.fq
26949

grep -c @DRR test.2.fq
26949

The dataset is here: https://github.com/macmanes-lab/Oyster_River_Protocol/tree/master/sampledata

I suspect some environment-specific issue here, but I can't understand what this could be.

[macmanes]

Further, when I reveres the order of the reads on input (incorrectly)

perl run_rcorrector.pl -1 test.2.fq -2 test.1.fq

I end up with fewer R1 reads.. Something about the -2 position/flag.

[mourisl]

Could you please try the newest version that has not yet been released yet? It could be fixed from the update of a2624ec from April 14, 2017.

[macmanes]

Yes, that fixes it. Are you willing to cut a patch release so I can get the fixed version in to bioconda? Thanks!

[mourisl]

Thanks! I just released the v1.0.3.

[mourisl]

I think this time I found the real reason causing the dropping reads from the second-mate file. And I've release v1.0.4 for this fix.