Identify gene modules correlated with INS secretion score and/or disease

[adamklie]

Context

For igvf-data/igvf_sc-islet_10X-Multiome. Previous attempts in scratch/2026_01_30/7_differential_analysis used DEGPatterns clustering + MAGMA enrichment. This was the basis for tools/gene-set-analysis.

A first-pass per-cell INS secretion score is available (Junxi — see #3, closed 2026-04-16). The score uses curated insulin-secretion gene sets + decoupler/AUCell. Iteration on the score itself (gene set refinement, alternate scoring methods, validation against experimental GSIS) now lives in this issue.

Tasks

• Coordinate with Luca and Sara to run LIGER on the data
• Explore other latent factor models (hdWGCNA?)
• Check if identified modules correlate with INS secretion score or disease status (e.g. MAGMA for T1D/T2D GWAS loci enrichment)
• Iterate on the INS secretion score (gene set refinement, alternate scoring methods, GSIS validation) — inherited from closed Impute INS secretion score at single cell level #3

[adamklie]

Dependencies

• Blocked by Impute INS secretion score at single cell level #3 (needs INS secretion score), Finalize DESeq2 pseudobulk methodology for RNA and ATAC #4 (needs finalized DESeq2 results)
• Feeds into Prioritize cREs for CRISPRi validation #16 (cRE prioritization)

[adamklie]

Status Update (2026-04-09)

Still blocked on #3 (collaborator, low priority for now) and #4 (DESeq2 finalization). No updates.

[adamklie]

Status Update (2026-04-16)

Still blocked on #4 (DESeq2 finalization) for module discovery at scale.

Scope expanded: closed #3 rolls in here. Junxi's first-pass per-cell INS secretion score is available; any further score iteration (gene-set refinement, scoring method swap, GSIS validation) happens under this issue rather than as a separate task.