Thus far, I have been looking at all the data found on this page: Multimodal Characterization in Mouse Visual Cortex - brain-map.org
for the combined morphology, electrophysiology, and transcriptome data for individual cells. I now have no problem (I think) looking at all 3 data types for a cell of my choice.
However, I was reading this paper “Single-neuron models linking electrophysiology, morphology, and transcriptomics across cortical cell types” that has used this data, or something similar, to model individual neurons. The problem is that that paper links to this site: Cell Features :: Allen Brain Atlas: Cell Types
for electrophysiology and morphology data and it seems to have completely different data from the first link I provided (and does not appear to support a convenient bulk download).
If I select any of the mouse neurons and take its ID, that number does not appear anywhere in the cell metadata or cell manifest spreadsheets that can be found in the first link I provided. Additionally, the paper claims to have reconstructed 230 cells but if you select ‘Has All-active Biophysical model’ you only get 107, even including the perisomatic models I think you only get up to 207 total which makes me think I am looking in the wrong place.
Simply put, where do I find combined MET datasets for cells that have an all active biophysical model? If this data is in the original link I provided, how am I meant to tell which cell is which?
The paper you’re referring to (Single-neuron models linking electrophysiology, morphology, and transcriptomics across cortical cell types) doesn’t actually use the data sets described at the first link you mentioned (about Patch-seq data sets from mouse visual cortex). It makes use of an earlier data set that is publicly available at Cell Features :: Allen Brain Atlas: Cell Types (described in the publication Gouwens et al. (2019)) and combines that with a transcriptomics-based study (Tasic et al. (2018)). The earlier data set has morphology and electrophysiology from the same cell, but no transcriptomics data (they were not Patch-seq experiments).
I believe the models described in that paper are also not the same as those published at Cell Features :: Allen Brain Atlas: Cell Types - I think the models described in that paper were developed later. It looks as if they are available at a GitHub repository for that paper. So that may explain the discrepancy in the number of models that you saw.
If you’re interested, the data in the Cell Types Database can be downloaded in bulk more easily using the Allen SDK.
You will have to forgive my unfamiliarity with the metadata here. When I look at HOF models, the mouse class data, or umap embeddings there is a given cell id. However, the files for the raw gene expression, exc_expression_all.csv for example, there is a sample id and no cell id. There must be metadata linking the two, no? I have been looking all through All-active-Manuscript/assets at master · AllenInstitute/All-active-Manuscript · GitHub and I am sure I am just missing it. If you could tell me how to connect the two I would very much appreciate it.
There won’t be any one-to-one correspondence between cell ID (for models) and sample ID (for single-cell RNA-seq data) because the cells used to build the models did not have any transcriptomic data collected from them - the paper you are looking at took two separate data sets and made comparisons between the two at the level of broad cell groupings (like Pvalb+ interneurons, or L5 PT cells). I believe the classifications of the models into those broad groups are in the Mouse_class_data.csv file (which has cell IDs as well as a Broad_Cre_line column). It looks like the files exc_ttype_expression.csv and inh_expression_filtered.csv have sample IDs (for the scRNA-seq cells) and subclass_label columns, which could be used to make those same kinds of comparisons at the broader group level.