MALT/MEGAN sample inconsistencies



I ran a bunch of ancient metagenome samples through MALT-X as

malt-build -i /. -s Protein -t 20 -d index -c Taxonomy

malt-run -i -d index -m BlastX -o . -a . -f Tab -oa . -t 20

and decided I wanted to look at the SEED classifications in the output. To do this, I uploaded the blast tab-formatted output file from MALT into MEGAN v6.1.8 with the acc2SEED-May2015.abin mapping file. When I compare the taxonomy profiles between the MALT-generated .rma6 files and the MEGAN-analyzed MALT tab-formatted alignment files, I see some big differences.

First, there are 4 samples that have very few assigned reads in the MEGAN-generated profile, but don’t have this problem with the MALT-generated profiles, and also then have very few SEED assignments.

Both MEGAN- and MALT-generated versions of the rma6 files have the same number of reads, but the % assigned in the MEGAN-generated files ranges from 2%-51%, with 86% of assigned reads only at the Kingdom level. The MALT-generated files for those same 4 samples have 99% of the reads assigned, and from 19-46% of assigned reads are only at the Kingdom level.

Second, in the remaining samples I see ~2000 species/sample in the MALT-generated profiles, vs ~300 species/sample in the MEGAN-generated files.

Is anyone able to explain why I see these discrepancies? Is there a way to make the MEGAN-generated profile more similar to the MALT-generated profile, so that the SEED profile is more similar to what it would be if it were part of the MALT-generated file?



Dear Irina,

my guess is that MALT and MEGAN are using different parameters for taxonomic and functional binning. Use the -v option with MALT to see all the options that program is using.
Alternatively, load the the rma file into MEGAN and then select the Options -> LCA Parameters menu item to display the parameters dialog and see what parameters were used by MALT to perform the analysis. Compare that to the parameters used for the BLAST tab file.
Are they the same? If so, please share the files with me and I will take a look at them.


Hi Daniel,

the LCA parameters were different so I adjusted them to be identical and the profiles now look the same.