How can i calculate the importance scores to generate a heatmap ?

[M_F (Meriem)]

Dear all,

I want to generate a heatmap but i should calculate the importance scores as required by the following script
qiime sample-classifier heatmap
--i-table ./dada2_table.qza
--i-importance ./sample-classifier-results/feature_importance.qza
--m-sample-metadata-file ./metadata.tsv
--m-sample-metadata-column genotype_and_donor_status
--p-group-samples
--p-feature-count 100
--o-heatmap ./sample-classifier-results/heatmap.qzv
--o-filtered-table ./sample-classifier-results/filtered-table.qza

Is there any command to use to generate the importance scores in a file called feature_importance.qza ? according to which criteria the value 100 was selected for feature-count ? is it compulsory to put the following options --p-group-samples
--p-feature-count 100 \ to run the command ?

Thank you

[llenzi (Luca)]

Hi @M_F,

the 'qiime sample-classifier heatmap' is the visualiser tool proposed for the sample classifier pipeline, the feature_importance.qza is the output for the 'qiime sample-classifier classify-samples' plug in. For the full example, please see Predicting sample metadata values with q2-sample-classifier — QIIME 2 2020.8.0 documentation

Hope it helps

[Nicholas_Bokulich]

@M_F,
To add to @llenzi's advice, if you just want a heatmap of all features in your data, see "qiime feature-table heatmap".

Good luck!

[M_F (Meriem)]

@llenzi and @Nicholas_Bokulich

thank you for the advises.

I run the command qiime feature-table heatmap for all the data and i obtained this message :
Plugin error from feature-table:

Unable to allocate 41.0 TiB for an array with shape (5637741168151,) and data type float64
i can't split the file because i need all those informations What can i do ?
best

[llenzi (Luca)]

Hi @M_F,

You could filter the low abundance ASVs across all samples to reduce the ASVs number (by using feature-table filter-features plug in filter-features: Filter features from table — QIIME 2 2020.8.0 documentation), or split the heatmap by sample groups, depend on what make more sense in your design.

Cheers

[M_F (Meriem)]

Hello @llenzi

Thanks for the response. Please can you tell me where can i find this information the frequency of features ?
best

[llenzi (Luca)]

Hi @M_F,

I suppose it depends on the data, I would try different frequency and see the results, what would make more sense. Start with a low number (eg 0.05 ) and see if you get the heatmap, if not try increasing the min freq.
Cheers