Running MEDI on Zymo mock community

[seethabala97-hub]

Thank you so much for developing MEDI. I found it as a cool tool to use. I ran MEDI on a set of stool metagenomes that included a Zymo mock community sample (13 million reads, ZymoBIOMICS® Microbial Community Standard), which was included in the sequencing run to account for plate to plate variation. the median sequencing depth for my gut microbiome samples ranges from 25-75 million.
I found out that the food_abundance.csv reported several food-associated taxa in the Zymo sample, including Artemisia (~11,633 reads), Juglans regia (~4,280 reads), Cucurbita moschata (~3,252 reads), Brassica oleracea, Vaccinium myrtillus, and Origanum. Have any other users observed this before? and how did you deal with detection limits and further filtration to obtain confidence in these findings.

[cdiener]

Hi, were those whole community standard or the DNA standard? Because most of the taxa in the mock are quite standard and should be part of the decoy database this may happen due to the following:

1. Incomplete MEDI DB built (was this one we shared?, what is the size of the hash.k2d file in the DB?)
2. Contamination of the plant reference genome with bacterial DNA (quite common, partially handled by MEDI but not always)
3. For whole community mocks note that the growth medium might contain DNA from other sources that may match bacteria (yeast extract, peptone digest etc).
4. Missing bacteria from the decoy set (unlikely, but one could check)

[seethabala97-hub]

Thank you @cdiener for your response. The zymo mock community that I had used was the whole community mock community rather than the DNA standard. The hash.k2d file in the MEDI database is approximately 496 GB. The species outputs from MEDI were able to identify all of the expected prokaryotic taxa, so those bacterial decoys are present in the MEDI database (Saccharomyces cerevisiae and Cryptococcus neoformans although present in the db, were not detected by MEDI genus outputs and were expectedly detected by MetaPhlAn).