DC1 anucleus is empty (n_obs = 0)

Competition Broad Institute Crunch #1
1

Hello,

I noticed that the anucleus in DC1 has 0 observations and no counts layer. Is this expected? The remaining files look ok on my end.

Thank you!

2

Hi,
Yes is expected.

3 
sdata['cell_id-group'].obs['group'].value_counts()

group
train         187577
test           17798
validation     10090<

So there are only 7 samples to process? In the description it is written (" sample: the tissue sample among the 8 samples to process." Crunch 1 – Oct 28 to Jan 31 – Predict gene expression | CrunchDAO Docs V3). This is also the only sample with train test validation split.

4

Each of the 8 samples DC1, DC5, UC1_I, UC1_NI, UC6_I, UC6_NI, UC7_I, UC9_I includes test and validation cells to predict

5
  1. If DC1 is empty (no train, no validation, no test), is it correct that we have access only to 7 samples? What is the role of DC1?
  2. I can see only “train” cells in DC5, UC1_I, UC1_NI, UC6_I, UC6_NI, UC7_I, UC9_I, but not validation and test. Is it correct?
6
  1. If DC1 is empty (no train, no validation, no test), is it correct that we have access only to 7 samples? What is the role of DC1?

There is no training data for DC1, as the objective is to predict at least one sample without any training data available for it.

  1. I can see only “train” cells in DC5, UC1_I, UC1_NI, UC6_I, UC6_NI, UC7_I, UC9_I, but not validation and test. Is it correct?

In sdata['cell_id-group'], you will find the train cells corresponding to the images as well as to the data in sdata['anucleus'], except for DC1. Additionally, sdata['cell_id-group'] contains the test and validation cells that you need to predict.