Issue with bacterial classification using Mothur v1.48.3 and SILVA v138.2

mothur bugs
1

Hello,

I would like to report an issue encountered with the classify.seqs command in Mothur version 1.48.3, using the SILVA v138.2 database (Full-length sequences and taxonomy references).

We use the ZymoBIOMICS Microbial Community DNA Standard (D6305) as a positive control !

I am working with 16S rRNA sequences targeting the V4 region, and I observed that sequences belonging to the genus Bacillus are often misclassified or remain unclassified. This behavior was not observed with the previous version of the database (v138).

By comparing the classification results obtained using SILVA v138.2 and RDP (PDS Version 19), we found that the PDS database (Version 19) is not a suitable choice for reliable bacterial classification.

Furthermore, using SILVA v138.2 with the KNN method failed to correctly identify three key bacterial genera: Escherichia coli, Salmonella, and Lactobacillus.
In addition, whether we apply a cutoff of 70 or not, Bacillus classification remains inaccurate with SILVA v138.2.

To address this issue, we opted to use an earlier version of SILVA (v138), adjusting the parameters with each run to improve classification accuracy.

Ultimately, we found that the SILVA v132 version, combined with a cutoff of 70, yields the best results for accurate and consistent bacterial classification.

image
image544×212 21.6 KB

Any suggestions would be greatly appreciated!

Thank you for your attention.

2

Hi, I’ll repeat what I posted to your question on the wiki’s issue tracker …

Thanks for your questions. First off, KNN is strongly discouraged. We recommend the default naive Bayesian classifier. I don’t know how to interpret your plots. What’s the y-axis - relative abundance? Are the plots from KNN or the naive Bayesian classifier? By cutoff are you talking about a percent similarity or confidence score? Are these data from your own sequencing data or the sequences provided by Zymo?

If you could post the fasta sequences for the positive control I can take a look.

If you can respond to these questions I can get you some answers.

Pat

3

For anyone following along I’ve posted a reply on the issue tracker: Issue with bacterial classification using Mothur v1.48.3 and SILVA v138.2 · Issue #139 · mothur/mothur.github.io · GitHub

4

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.