Open mortonjt opened 5 years ago
Thanks @mortonjt, this is a good idea. I agree that the names of the inputs could be more generic. For now, it is way too specific to the sample-vs-feature
case. However, instead of going the route of having a multiomics-biplot
, I would like to fix the biplot
command to use a better naming convention. If we think about it from the perspective of q2cli, and following your recommendation it would look like this:
# For the feature vs feature (mmvec) case
qiime emperor biplot \
--m-point-metadata metabolite-annotations.tsv \
--m-arrow-metadata microbe-metadata.tsv \
--i-ordination mmvec.results.qza \
--o-visualization biplot.qzv
# for the sample vs feature case
qiime emperor biplot \
--m-point-metadata sample-metadata.tsv \
--m-arrow-metadata feature-metadata.tsv \
--i-ordination pcoa-biplot.results.qza \
--o-visualization biplot.qzv
The code itself could check to ensure that there's an overlap between the point metadata and the points in the ordination object, as well as checking for the arrow metadata and the arrows in the ordination. I think that looks fairly straight forward. What do you think? Maybe we could refine the point
and arrow
names?
Another thing that might help is to check and raise a meaningful error if users accidentally inverted the order of the metadata. This could be checked based on an identifier match between the metadata files. For example: The point and arrow metadata seem to be in an incorrect order. The point metadata matches the arrows in your ordination, and the arrow metadata matches the points in your ordination.
Or something to that effect.
I think renaming to --m-point-metadata
and --m-arrow-metadata
makes a lot of sense.
And checking for the point / arrow metadata match is a good idea!
There has been a bit of confusion on how to generate multiomics biplots using mmvec.
This mainly boils down to the usage of
sample-metadata-file
andfeature-metadata-file
- this can get confusing with multiomics sincesample-metadata-file
is typically another type offeature-metadata-file
.The easiest solution that I can think of is to introduce another command called
multiomics-biplot
and rename the existing biplot command tosample-biplot
. In this way,multiomics-biplot
can be typed so that it can accept twofeature-metadata-file
arguments. Furthermore, we can change the argument names to something likepoint-feature-metadata-file
andarrow-feature-metadata-file
to reduce the confusion when trying to use this.Thoughts? Also will be happy to contribute.