Missing input gene volumes

[francesshei]

combine_markers_ccfv2_config.yaml requires various gene volumes as input (i.e.: data/ccfv2/marker_volumes/CNP.nrrd) but no instructions are provided on how to obtain these files. Should these files be curled, or are these files part of a missing pipeline step?

[mgeplf]

I think @drodarie should be able to explain that best.

[drodarie]

These datasets should be obtained through the DeepAtlas pipeline as describe here.
This is not part of the atlas-densities pipeline.

[francesshei]

How are multiple IDs for different genes supposed to be merged? I.e., there are multiple experiments for each gene listed in the yaml file . Should all of these experiments be interpolated and then averaged?
Is it "safe" to combine coronal and sagittal experiments? Sagittal experiments only cover one hemisphere (namely, the left hemisphere).

[drodarie]

In the README. It is mention that you can read Eroe et al. 2018 for more info on the method to combine. Alternatively, you can just read the docs of the function combine that is called at this step.

atlas-densities expect one 3D voxelized ISH dataset per gene. The operations of interpolation and merging necessary to obtain them are not done here. If you wish to replicate the results of Erö or Rodarie then this is the list of experiment ids used:

• mbp: 112202838,
• gfap: 79591671,
• s100b; 79591593,
• aldh1l1: 68631984,
• tmem119: 68161453,

For these papers, the ISH datasets were interpolated, and the sagittal experiments were mirrored. I attempted a combination of different datasets in my paper and it seemed to improve the results.