scPCIP: cross-species annotation of plant cell types with conserved gene programs

scPCIP learns conserved gene programs from 112,905 author-labelled plant cells and annotates cell types in species absent from training, outperforming marker-based methods across 26 held-out datasets.

• Why it matters: Cross-species marker genes often fail to transfer, limiting annotation of plant cell atlases when new species are added.
• What they did: The method trains ortholog-group gene programs for root, leaf, flower, and seed cell types, then applies them without a reference to new species using scPCIP.
• The result: In 26 held-out datasets, scPCIP outperforms marker-based and plant deep-learning approaches, matches supervised classifiers on the same atlas, and transfers to an independent study at lower accuracy.

AI summary of the abstract · may contain errors · not medical advice · read the paper.