Training feature classifier plugin error

[Natalia_Bednarska]

Hi,

I encountered error while training my classifier on V3V4 region using Silva_132_99_16S and reference-taxonomy-7-levels.

Plugin error from feature-classifier:

Unable to allocate array with shape (58095, 8192) and data type float64

Debug info has been saved to /tmp/qiime2-q2cli-err-n27inpo6.log

Any ideas what could have gone wrong?

Cheers

[Nicholas_Bokulich]

Hi @Natalia_Bednarska,
This is essentially saying your database is too large to process on your computer. What SILVA files are you using? You should use the 99% OTUs if possible — when I have seen this error in the past is usually when someone is training the database on the full, redundant dataset.

[Natalia_Bednarska]

Hi Nicholas,

Many thanks for your quick reply. I have used the files downloaded from SILVA132 package. I also trained my classifiers on 99%OTUs from Grene genes database, but got many redundant classifications for my basic standard microbiome sample.Hence I thought to try different classifiers. Now I am unsure if this is due to an error in selection of classifier or perhaps demultiplexing parameters ( I included region of 165 base pairs).

[Nicholas_Bokulich]

@Natalia_Bednarska

got many redundant classifications for my basic standard microbiome sample

many ASVs can and will classify to the same taxonomic label — just because the sequences are unique does not mean that they come from different species.

@Natalia_Bednarska

Now I am unsure if this is due to an error in selection of classifier or perhaps demultiplexing parameters

As far as I can tell, this is due to memory constraints on your computer. This should not have anything to do with demultiplexing, since you are running this on the reference data not your own data.