search

biomap-research / scFoundation

Apache License 2.0
236 stars 37 forks source link

Question for the number reported for the CDR task #6

Closed KatarinaYuan closed 10 months ago

KatarinaYuan commented 10 months ago

Hi, Thank you for the great work! And I'd like to kidnly ask for your help with the CDR task.

There seems to be a mismatch between the reported number (PCC above 0.93) in the paper and the expected PCC around 0.87 in the repo (https://github.com/biomap-research/scFoundation/tree/main/DeepCDR)

Could you provide more details on the evaluation setting? Do you have any suggestion on how we could run experiment to get the reported number?

Thank you for the help:)

WhirlFirst commented 10 months ago

Hi, The results showed in this repo are calculated on all test data, consisting of multiple drugs and cancer types. The number 0.93 reported in the paper demonstrated that the embedding-based model can achieve high PCC values in two cases, as shown in Fig. 3C. In other words, all these numbers are based on the same results, but with different sample sets of calculating PCC.

Thank you for your question. We will consider revising the description to make it clearer.