search

calico / solo

software to detect doublets
MIT License
85 stars 13 forks source link

Error using h5ad file #78

Closed jddavies95 closed 1 year ago

jddavies95 commented 1 year ago

Hey, I keep getting this error about the object not having the attribute 'loc'. I am using a h5ad file of unnormalized, filtered data exported from scanpy. Do you know what may be causing this? Thanks!

2022-11-23T16:57:41.248186372Z Traceback (most recent call last): 2022-11-23T16:57:41.248381225Z File "/opt/conda/bin/solo", line 33, in 2022-11-23T16:57:41.249388349Z sys.exit(load_entry_point('solo-sc', 'console_scripts', 'solo')()) 2022-11-23T16:57:41.249686579Z File "/opt/solo/solo/solo.py", line 319, in main 2022-11-23T16:57:41.249926121Z doublet_score = softmax_predictions.loc[:, "doublet"] 2022-11-23T16:57:41.252468271Z AttributeError: 'numpy.ndarray' object has no attribute 'loc'

davek44 commented 1 year ago

Hi, it's confusing for softmax_predictions to be an ndarray rather than a pandas dataframe. You can see here scVI SOLO predict returns a dataframe here: https://github.com/scverse/scvi-tools/blob/main/scvi/external/solo/_model.py#L392. Can you use the python debugger and try to figure out where softmax_predictions becomes an array for you?

njbernstein commented 1 year ago

@davek44 close this one too cause we didn't get a follow up?