Is my command right for classifying ? (I changed my classifier command to use q2-feature-classifier)

General Discussion
, , ,
1

I have used qiime2 with my own database made from NCBI database for 18s rRNA sequence.

However, I found that some of the taxonomic assigned results from my command was slightly different from the NCBI blastn results.

While I found the reason, I noticed q2-feature-classifier is replacing the plugin feature-classifier classify-consensus-blast nowadays.

Hence, I have tried to change my original command for q2-feature-classifier using my own 18s rRNA database.

From the original script, is the changed script for the new classifier right ?

(Original command)
qiime feature-classifier classify-consensus-blast
--i-query Pigeon18S_min1000/representative_sequences.qza
--i-reference-reads ncbi-refseqs_18S_v220513-seqs-derep-uniq.qza
--i-reference-taxonomy ncbi-refseqs_18S_v220513-tax-derep-uniq.qza
--p-maxaccepts 1
--p-perc-identity 0.95
--p-min-consensus 0.75
--p-query-cov 0.5
--p-evalue 0.001
--o-classification Pigeon18S_min1000/taxonomy_consensus_blast-asv.qza

(The new one) - first classifier making

qiime feature-classifier fit-classifier-naive-bayes
--i-reference-reads ncbi-refseqs_18S_v220513-seqs-derep-uniq.qza
--i-reference-taxonomy ncbi-refseqs_18S_v220513-tax-derep-uniq.qza
--o-classifier your_trained_classifier.qza

(second) Using the trained classifier.

qiime feature-classifier classify-sklearn
--i-classifier your_trained_classifier.qza
--i-reads Pigeon18S_min1000/representative_sequences.qza
--o-classification Pigeon18S_min1000/taxonomy_q2-classifier-asv.qza

Is this way right ? I want to use my own collected data for my taxonomy assignment.

Please teach me the right way for using q2-feature-classifier plug-in.

Thank you in advance for your help.

3

Hi @SingeunOh,

Hope you are well!

If you have a look around the forum there are posts that discuss this. For example this discussion is really useful. Also take a look at this paper.

I hope that's helpful,

Vic

4 Likes
4

I used the qiime feature-classifier classify-consensus-blast command to classify my own 18s rRNA sequence database obtained from NCBI. However, upon comparing the classification results with the NCBI blastn results for the same sequence, I noticed slight differences. I'm curious about the algorithmic variations between the classification method employed by the plugin and the NCBI blast.

For instance, here's an example sequence: ATCGCTGCAACCGATCGGAGGGTCCTGTGAACTCATCGGACTGACTAGCTCCGCCTTGCGCGAGGCTGGTCGGGAAGTTGCGTAAATAGAGCCCTCTAAAGGATGCAAAAGTCGTAACACGGTTTCCGTAGGTGAACC

The top match from the classifier was identified as Eimeria. Could someone explain the differences in the classification algorithms between the QIIME2 plugin and NCBI blast? I'd greatly appreciate insights or suggestions for further analysis.

  • this is the scripts that I have used

qiime feature-classifier classify-consensus-blast
--i-query Pigeon18S_min1000/representative_sequences.qza
--i-reference-reads ncbi-refseqs_18S_v220513-seqs-derep-uniq.qza
--i-reference-taxonomy ncbi-refseqs_18S_v220513-tax-derep-uniq.qza
--p-maxaccepts 1
--p-perc-identity 0.95
--p-min-consensus 0.75
--p-query-cov 0.5
--p-evalue 0.001
--o-classification Pigeon18S_min1000/taxonomy_consensus_blast-asv.qza

5

Hi @SingeunOh,
I think this video address your question perfectly: https://www.youtube.com/watch?v=Z9w2VZHJMZs&list=PLbVDKwGpb3XmkQmoBy1wh3QfWlWdn_pTT&index=15

I hope that helps!

4 Likes
7

I will definitely try to study the video and review for my better understanding ! Thank you.