Discrepancy between metaphlan3 and Humann3 in profiling micriobiota


When using metaphlan3 and humann3, I found there was a discrepancy between metaphlan3 and Humann3 in profiling microbiota.

The versions of metaphlan3 and humann3 are “MetaPhlAn version 3.0 (20 Mar 2020)” and “humann v3.0.0.alpha.4” respectively.

My code for metaphlan3 is

nohup metaphlan sub_M1_35000000.fq --nproc 4 --input_type fastq --no_map -o profiled_metagenome_M1.txt > log_M1 2>&1 &

Code for humann3 is

nohup humann -i sub_M1_35000000.fq.gz -o ./ --threads 8 --output-basename M1 --diamond ~/toolkit/diamond --remove-temp-output > log_M1 2>&1 &

I noticed that the annotated species from humann3 are more than from metaphlan3. And I found that some differential genes or pathways were assigned to the species that were absent in the results of metaphlan3. So I have no idea what happened. How can I interpret the result of humann3? Or it there anything I misunderstood?

Thanks in advance!

Hi - Thanks for the detailed post! HUMAnN v3 calls MetaPhlAn v3 for the taxonomic profile step. You can bypass this step by providing the MetaPhlAn profile as input when running HUMAnN. Because HUMAnN relies on MetaPhlAn for taxonomic profiling the species identified for each run should be the same. Is it possible when you are running HUMAnN it is picking up a different version of MetaPhlAn then when you are running MetaPhlAn directly?

Thank you,

I didn’t change the metaphlan when running humann3. Is it possible that it is due to the database of metaphlan3. I noticed that metaphlan3 would check the database and automatically install it. How often the database update, and if it has been updated automatically, would the microbiota profile be distinct with the previous?

Hello, Changes to the MetaPhlAn database would possibly change the results. The database changes with the version of MetaPhlAn. There were a few database iterations during the development of MetaPhlAn v3. The MetaPhlAn database version is usually printed to the comments at the beginning of the taxonomic profile. If you check out this file you can see if the same database was used for both runs.

Thank you,

But the database is identical between two runs, that is “mpa_v30_CHOCOPhlAn_201901”. The first run was about Apr 8, the second run was performed recently. And I found the metaphlan databases appeared to update on Apr 14. Are there some different versions of “mpa_v30_CHOCOPhlAn_201901”, although they have the same ID.

And could you kindly provide the old database of the metaphlan3 of 1st run, which should be before Apr 8. I need to get a consistent analysis. Thanks very much!!!

Hi, there are no different versions of the v30 database, it was re-uploaded due after fixing some genomes accession that were duplicated, but this was done only in the .pkl file and the fasta remained untouched.

Can you check if the same MetaPhlAn version is called by HUMAnN? Also, can you upload here the two bowtie2out files?


Unfortunately, I didn’t keep the intermediate files. I installed the metaphlan3 by Conda. The 1st run of metaphlan3 was Arip 8, the second run of metaphlan3 was Sept 10. The results of the two runs are different. Besides, I noticed that my present metaphlan3 was installed on Arip 13, whose version is MetaPhlAn version 3.0 (20 Mar 2020). It means the version of metaphlan3 of two runs is different. If the database remains the same, is it due to the different versions of metaphaln3? But I cannot recall the exact version of 1st run.

Here is the link of metaphlan3 in conda - https://anaconda.org/bioconda/metaphlan/files.

There are serval versions of metaphlan3.