Add Argelaguet's gastrulation scNMT

[imallona]

Evaluate "normal" regions, e.g. proms/ K4me3, polycomb etc across gastrulation.

https://pmc.ncbi.nlm.nih.gov/articles/PMC6924995
https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE121708

Data are downloaded by #20 at https://github.com/imallona/yamet/blob/simulations/simulations/Snakefile

It's mm10.

[imallona]

Check enhancers; timecourse; lineage. Discard TET KOs (?).

Genomic compartments and annotations hints from past papers:

• scmet:
  • "Here, we used scMET to quantify DNAm variability at promoter elements, which we subsequently contrasted to RNA expression heterogeneity for the corresponding genes (...) When comparing residual overdispersion estimates for RNA expression and promoter DNAm, there was no clear genome-wide association (Fig. 4a). However, when restricting to genes that display high levels of transcriptional variability, two main groups can be identified."
  • so perhaps typical polycomb and marks, instead of gene-centric slices?
• epichaos:
  • "we first applied epiCHAOS to single-cell nucleosome, methylation, and transcription sequencing (scNMT-seq) DNA methylation data from mouse gastrulation [42]. Here we observed the highest epigenetic heterogeneity in the epiblast compared to more differentiated germ layers (Fig. 5A), as well as a moderate correlation between promoter-wide DNA methylation-based and ATAC-based epiCHAOS scores from the same cells (Fig. 5B)."
  • so perhaps typical by-date, by-layer stratification? worth applying to atac for some definition of binarized peaks?
• original scnmt gastrulation paper:
  • "We next asked how the epigenomic patterns associated with germ-layer specification arise during development. DNA methylation levels in endoderm and mesoderm-defining enhancers follow the genome-wide dynamics, increasing from an average of 25% to 80% in all cell types (Figure 3 and Extended Data Fig. 9). Upon lineage specification, they undergo concerted demethylation to ~50% in a cell type specific manner. The opposite pattern is observed for chromatin accessibility; accessibility of meso- and endoderm-defining enhancers initially decreases from ~40% to ~30% (following the genome-wide dynamics) before becoming more accessible (~45%) upon lineage specification."
  • so perhaps focusing on enhancers vs promoters, along layers and timecourse?