Z-Screen Source Data (June 2026 Release): single-nanowell transcriptomics, image embeddings, and compound chemistry

Curated public source-data layers for the Z-Screen platform, a modality-agnostic single-nanowell drug-discovery screen. Includes raw transcriptomic counts (768,162 nanowells x 46,944 genes) with a 1-to-1 master well table, normalized image (morphology) embeddings and biomarker intensities, a public compound/building-block chemistry feature table, an ActiveSeq same-well RNA+imaging calibration layer, and an external-reference benchmark surface of ranked gene-list signatures (Replogle, Norman, scPerturb-THP1, LINCS L1000, Tahoe-100M) plus gene sets. See README.md and MANIFEST.csv (SHA256 + row/column counts) in the record. Raw external H5AD matrices and manuscript/working directories are intentionally excluded; the redundant plain same-well join is omitted in favor of the public-compound-mapped superset.