Why do you ignore the 0s?

[many-kalin]

Hi, I noticed that you are ignoring zero counts when computing the metrics competitions/competitions/broad-2/scoring/scoring.py at f74a00d072730d1c8b4c1bd567d4893adbf2f18c · crunchdao/competitions · GitHub, and I was wondering about the rationale behind this choice. Zero values (indicating that a gene is not expressed or captured) are biologically meaningful and can provide important insights into gene regulation, dropout events, or technical artifacts. Excluding them might bias the analysis or overlook crucial patterns in the data. Embracing the dropouts in single-cell RNA-seq analysis | Nature Communications

In crunch 1 you were not ignoring the 0s…

[enzo]

We could not get a 100% correlation when comparing the ground truth to itself. So we filtered out the zeros.

After an analysis where we recalculated all the predictions keeping the zeros, the ranks are almost exactly the same.