Bug with n_metagenome_reads

[jeanmng]

Hi,

I hope you are doing well, Thank you fro your work on metaphan3. I think I found a bug

I use metaphan3 with the --nreads, --unknown_estimation and --input_type sam, I have the following error message:

fraction_mapped_reads = min(tot_nreads/float(n_metagenome_reads), 1.0) if ESTIMATE_UNK else 1.0
  TypeError: float() argument must be a string or a number, not 'NoneType'

I did a bit of digging in the code and found that the

• n_metagenome_reads variable is instantiated by the map2bbh function
• this function instantiate an internal n_metagenome_reads variable with None (https://github.com/biobakery/MetaPhlAn/blob/446ee61b05a5063365a315ddba8c54bf45fcae66/metaphlan/metaphlan.py#L796) but doesn’t change the at any point when input_type == 'sam' https://github.com/biobakery/MetaPhlAn/blob/446ee61b05a5063365a315ddba8c54bf45fcae66/metaphlan/metaphlan.py#L804-L813. So the returned value is indeed None.
• later here https://github.com/biobakery/MetaPhlAn/blob/446ee61b05a5063365a315ddba8c54bf45fcae66/metaphlan/metaphlan.py#L1003-L1009 there is a check on n_metagenome_reads (None) and pars['nreads'] (the value I gave via the command line) but it only exit with an error (which is doesn’t because ars[‘nreads’]` has a value
• Come https://github.com/biobakery/MetaPhlAn/blob/446ee61b05a5063365a315ddba8c54bf45fcae66/metaphlan/metaphlan.py#L1059-L1060, where the error happens, n_metagenome_reads is still None, then it fails

Maybe pars[‘nreads’] should be used instead. This issue doesn’t occur when I omit the --unknown_estimation tag

[fbeghini]

Thank you for spotting this, I may have forgotten to specify to use the pars[‘nreads’] value when a SAM file is used as input. The latest commit fixes this.

[jeanmng]

Thank you very much! That was quick.

Hopefully the conda/singularity/docker container will be updated soon.