* SCP-2611 Correct gene models for dense matrices and account for single column feature files #145
Conversation
Codecov Report
@@ Coverage Diff @@
## development #145 +/- ##
===============================================
+ Coverage 68.83% 69.00% +0.17%
===============================================
Files 22 22
Lines 2663 2678 +15
===============================================
+ Hits 1833 1848 +15
Misses 830 830
Continue to review full report at Codecov.
|
knapii-developments
requested review from
devonbush and
eweitz
and removed request for
eweitz
knapii-developments
changed the title
eweitz
changed the title
devonbush
left a comment
devonbush
left a comment
There was a problem hiding this comment.
Code looks good. Please add a test case that confirms gene models get created for genes with no non-zero expression scores (i.e. a test case that would fail without this fix)
|
knapii-developments
changed the title
devonbush
left a comment
devonbush
left a comment
There was a problem hiding this comment.
Needs a test to cover the single-column gene file case. And one minor code format suggestion
Co-authored-by: Devon <dbush@broadinstitute.org>
| # models maybe less than the batch size. | ||
| if len(gene_models) > 0: | ||
| self.create_models( | ||
| [], [], None, None, gene_models, data_arrays, num_processed, True |
There was a problem hiding this comment.
Thanks for taking the time to explain this to me, Eno!
There was a regression for parsing dense matrices where gene models were only created if the gene had expression data. The PR addresses this issue by creating gene models regardless of the presence of expression data.
There was also a second regression. The feature file in an MTX bundle can have 1 column. As stated by 10X genomics, we now set the gene name to the gene ID in the event 1 column is present.