Benchmarking biological knowledge-graph priors against single-cell foundation models for Perturb-seq prediction

Ni et al. (Southeast University; Computational Biology and Chemistry 2026) compare STRING/Gene Ontology graph priors with dense, self-edge-only and degree-rewired supports in one fixed graph-attention backbone on four public Perturb-seq datasets (Adamson, Norman, Replogle K562 and RPE1), with GEARS, scGPT and Geneformer evaluated on the same held-out guide conditions. At 200 highly variable genes, STRING-GO improved Pearson r by +0.016 (95% interval +0.010 to +0.022) over dense support; the effect fell to +0.009 at 1000 genes. Of 36 matched contrasts with external models, 32 were within a ±0.030 envelope and four favoured the external model. The authors state the main comparison was formulated during peer-review revision and that the modest gain does not by itself establish biological relevance.

Authors / organisation

Ni J, Zhang X, Li M, Cai Y, Xie Z, Liu Y.

Venue

Computational Biology and Chemistry

Date

2026-10-09

Type

paper

Year

2026

DOI

10.1016/j.compbiolchem.2026.109480

PMID

42860076

Verification

PubMed E-utilities PMID 42860076 (title, 6 authors, Southeast University affiliations, Comput Biol Chem 126(Pt 2):109480, epub 2026-10-09, abstract) + Crossref (journal-article, issue date 2027-02) fetched 2026-10-20; full text not read

Benchmark & validationPerturb-seqFoundation modelPerturbation

Open in the searchable library → · JSON