Issue reproducing TTN detection counts in In Silico Treatment Analysis

#579
by LucasACyyy - opened

Hi,

I’m reproducing the In Silico Treatment Analysis using the SCP1303 dataset. My cell type classification accuracy reached 0.8385, which aligns with your results.

However, I'm seeing a discrepancy in the perturbation step. I filtered the 29 training samples for disease == "NF" and cell_type == ["Cardiomyocyte_I", "Cardiomyocyte_II", "Cardiomyocyte_III"], resulting in 33,788 cells. While TTN is highly expressed across almost all these cells in my run, your reproduction table shows its n_detections at only around 10,000.

Did you apply additional quality control or specific filtering criteria for the cells used in this analysis?

Best,

Sign up or log in to comment