Replicating Fig 3c — request for lmplots source from helper_functions.py

Technical
,
1

Dear all,

I am writing for clarification on the implementation of the lmplots function used to generate Figure 3c of Gabitto et al. 2024 (Nature Neuroscience).

In the context of a project on a subpopulation in MTG, I have been replicating the trajectory analysis using your publicly available SEA-AD_2024 code. The qualitative shape of my reproduced curves matches the figure, but absolute log(relative abundance) values are systematically offset by approximately 0.4 to 0.7 log units. For example, my Sst subclass trajectory spans -3.0 to -3.5 versus the published -2.4 to -3.0. I have tested every variation I could infer from the legend description, including library-vs-donor aggregation, all supertypes vs only those in affected_supertypes, and several pseudocount placements, without closing the gap.

The helper_functions.py referenced by the Figure 3c notebook does not appear in the public repository. Would it be possible to share:

  • The body of lmplots (specifically the section handling y_scale="log_relative"), or
  • A brief description of the exact normalization formula used?

A pointer to any related documentation I may have missed would also be very welcome.

Thank you very much for your time. The SEA-AD resource has been invaluable for my work.

Best,

ag

2

Hi @granzota ,

Let’s see if we can figure out where the discrepancy is coming from! The lmplots code is here on our git repo: SEA-AD_2024/Figures/helper_functions.py at main · AllenInstitute/SEA-AD_2024 · GitHub. Are you able to access it? If so, maybe give that a try.

Best, Kyle

3

Hi @kyle.travaglini ,

thanks for the lmplots code. It works nicely now. The gist is that the trasform is per-supertype and the lines are pooled and not summed. My bad. thanks for your help.

Best,

ag