Cross-cohort metadata harmonization
#2
by Shicheng - opened
Packaging nearly 29 million brain cells as row-sharded Zarr stores reduces a roughly 7.13-terabyte dense matrix to practical chunked access, which could substantially lower barriers to atlas-scale analysis. How are you thinking about harmonizing donor, disease, and assay metadata so these stores can support reproducible cross-cohort target discovery?