Can't reproduce Figure 3e

[Lulu0015]

Hi,

Thank you for developing this fantastic tool! I am trying to reproduce Figure 3e. I used the "parallel_inference.py" script from the /scripts/inference_motif_deletions folder to predict the motif effect after mutating all matching sites for GATA1 (motif #10 in the motif table). Using the code in /notebooks/figure3/Figure3.ipynb, I successfully reproduced the predictions for the reference sequences. However, my results for the mutated sequences differ from the example in the notebook. Do you have any suggestions for debugging this issue?

Thanks so much!

[lauradmartens]

Hi!

Good catch, there was indeed a small mistake in the script that meant that the model never saw the alternative sequences. We fixed it now, let us know if it works for you!

Best,

Laura

[Lulu0015]

Hi Laura,

Thanks so much! It was helpful. I can reproduce the motif effect for acc. now, but the motif effect for expr. still looks different. Any insight?

Thanks,
Lu

[lauradmartens]

Hi Lulu,

It should now be fixed, sorry for that. After our refactoring, we accidentally had incoherent settings for reference and motif deleted sequences. We also increased the cache_size argument which should lead to a significant speedup for the motif deletions.

Hope that helps!

Best,
Laura & Johannes