What problem does it solve? When scRNA-seq and scATAC-seq come from separate cell pools, there are no shared barcodes, so cell identity cannot be read directly off ATAC cells and peak calling cannot be grouped by cell type. This Skill bridges the modalities with a gene-activity matrix and staged CCA label transfer, then calls, merges, tiers, and validates a consensus peak atlas. ## Core Features & Use Cases - Staged label transfer: Transfers coarse lineage labels (R1) then refined subtype labels (R2) from an annotated RNA reference onto ATAC cells via gene-activity space, excluding low-confidence cells from label-resolved peak calling. - Three complementary peak-calling strategies: Calls peaks per coarse cell type (A), per refined subtype × condition (B), and label-free per condition (C) as a transfer-error-robust hedge, then merges them with support voting. - External consensus and tier stratification: Folds published ATAC datasets into a consensus peak source with a 0.25 support threshold and bias gate, and stratifies peaks into priority tiers including protected rare-biology Tier 0 (0a/0b/0c). - Use Case: Given separate scRNA and scATAC experiments on the same immune system, transfer cell-type labels onto ATAC cells, call peaks three ways, gate rare single-strategy peaks on an external-correlation test (r >= 0.4), and produce a validated consensus atlas. ## Quick Start Use the peak-atlas-unpaired skill to transfer labels from my annotated scRNA reference onto my scATAC object and build a tiered consensus peak atlas across conditions.
