Resolve the conflicting K562 cell-identity annotation

Determine whether the A549 value recorded in the upstream source-cell-type metadata is a metadata error or represents a distinct source-specific field, given the conflicting K562 cell-identity annotation in the internal contract.

Background

The K562 genetic perturbation benchmark uses the Norman 2019 CRISPR activation dataset and is represented internally with both a cell identity identifier and an upstream source-cell-type annotation. These two metadata fields conflict: the contract identifies the benchmark as K562, whereas the upstream field contains A549.

The manuscript explicitly leaves unresolved whether the A549 annotation is erroneous or carries a different source-specific meaning. Resolving this issue is necessary to establish the provenance and biological identity of the benchmark used for model evaluation.

References

The current internal contract retains both cell_identity_id=CVCL_0004 and the upstream field source_cell_type_value=A549. Because these annotations conflict with the stated K562 benchmark identity, this provenance inconsistency is retained explicitly pending confirmation of whether the A549 annotation represents an upstream metadata error or another source-specific field.

— PRAXIS-VirtualCell: A Programmable and Trustworthy Framework for Agentic Virtual Cell Experiments  (2609.29221 - Ma et al., 24 Sep 2026) in Methods, K562 genetic perturbation benchmark