Hi, I have a quick question about the meaning of my submission ranking. I can only see the ranking based on my submission id. Is it possible to see the performance based on ranking? Thanks.
Also, I am a bit confused by the input data of chrunch 2:
It seems that some datasets have zero cells, while others have different number of anucelus and cell_id_elements. Is it normal? My understanding of imputation problem is that we must at least have gene expression of certain anchor genes.
Furthermore, I tired my own method but met one error: forbidden library: xxx.
Does this mean we can only implement our own methods without importing other library? That is wired. Thanks.
Libraries are restricted to prevent abuse.
Have a look at the documentation: Whitelisted Libraries | CrunchDAO Docs V3
If you want to request a library, don’t hesitate to send me a message.
Hi, only DC1 doesn’t have any gene expressions, and this is expected. You can find more details here: DC1 anucleus is empty (n_obs = 0)
Thanks a lot. I will send you a message.
Hi, thanks a lot. Therefore, will DC1 be used in the second challenge or not? If we intend to impute DC1, I think we need to predict the cell first and then run the imputation algorithm.
Furthermore, I found many cells are missing in the sanxelus layer, and these cells exist in testing and validation set. How to interpret the result? Do we need to take care of the missing parts or we only need to try to predict the resting genes of measured part?
The data present in anucleus are the ones you can use for training. It is normal that no validation or test cells are found in anucleus, as these are the cells to be predicted.
Hi, thanks for your reply. I do not think it is the same as the definition of imputation. See https://www.nature.com/articles/s41592-022-01481-8, imputation always has anchor genes as references.