Qemistree updates

[anupriyatripathi]

The following features need to be pushed into the new GNPS release:

1. Adding the option to select the instrument the data was run on for FP prediction accuracy
2. Adding the option to use euclidean or jaccard distances to cluster fingerprints
3. Running Classyfire on smiles from ms/ms matches or csi fingerID predictions
4. Figuring out how to render the tree in a user-friendly manner

[mwang87]

Can you provide the exact commandlines we would be using to run this?

[anupriyatripathi]

I've sent you an email with the commands for this.

[anupriyatripathi]

Test tasks:

Name	Description	FBMN	Qemistree
QE QERT C18 benchmarking	QE example	Job	Job
Western vs. rural skin samples	QTOF example	Job	Job
Mice serum samples	QTOF no metadata example	Job	Job
Fungal gardens	QTOF example	Job	Job

[anupriyatripathi]

@mwang87 The current workflow renders a file that appears to be an output of one of the early steps in qemistree workflow i.e. output from SIRIUS. We should instead be rendering the feature data file produced from the last step of the processing i.e. the output of get-classyfire-taxonomy which is a .qza file that can be exported to a .tsv as follows:

qiime tools export --input-path classyfire_results.qza --output-path classyfire-results

classyfire-results is a folder that would contain a .tsv file called feature_data.tsv. We should render feature_data.tsv

[mwang87]

I agree they may not be correct, but please check the output files. Are they correct for all the tests you have run? Lets take this one step at a time.

[anupriyatripathi]

@mwang87 The output files look correct; just highlighting the TBDs here. Upon further investigation, it looks a file called formula.qza is being rendered currently.

[mwang87]

Sounds good. I'll pull out that file. What are the contents of that feature_data.tsv file? Which columns are meaningful to display?

[anupriyatripathi]

The classified feature data has the following columns. I don't think there's an issue with displaying everything. table_number and is not relevant though as we would not be supporting meta-analyses for the time being. And the column smiles is just either ms2_smiles or csi_smiles (when ms2 _smiles in NaN)

id	#featureID	csi_smiles	ms2_smiles	ms2_compound	ms2_adduct	table_number	smiles	annotation_type	kingdom	superclass	class	subclass	direct_parent

[mwang87]

Closing as the latest changes seem to be good. Will go out in release 18