question · intermediate · partly written
What fraction of a graph benchmark score survives a degree-only baseline and a by-disease split?
Nobody outside the builders has reported how much of a PrimeKG or ogbl-biokg link-prediction score is left after a degree-only baseline and a split by disease.
What is known more
A degree-only edge prior reached AUROC of at least 0.95 on 17 of 20 Hetionet-derived networks (best 0.9971, worst 0.7697) and about 0.5 on a separate PPI network. TransE scored high-degree gene answers about ten times better on MRR than low-degree ones (0.053 against 0.006). Both studies were done by groups close to the graphs or to industry users, and no independent audit of Hetionet or PrimeKG leakage was found. 12
Reported, not yet replicated
Why it is still open more
Hetionet's degree-weighted path counts and permuted networks were designed against degree bias, but whether they remove it fully was not tested in the sources opened. TxGNN's own comparison of standard and zero-shot splits gives relative gains of 4.3% and 19.0% against different baselines, which is not an absolute drop. 34
Reported, not yet replicated
Still open here
- What does a degree-only prior score on PrimeKG drug-protein edges? No degree baseline for PrimeKG was found.
- Does the Rephetio model (AUROC 85.5% on DrugCentral indications) survive negatives matched on compound and disease degree? No published run was found.
Not yet written: Observation, Hypotheses, What is missing, What is known, What is not, What would settle it, Who works on it.
Linked from
Sources
- Zietz M, Himmelstein DS, Greene CS, et al.. The probability of edge existence due to node degree: a baseline for network-based predictions. GigaScience (2024). paper; opened 2026-11-18 all records citing this
- Bonner S, et al.. Implications of topological imbalance for representation learning on biomedical knowledge graphs. arXiv 2112.06567 (2022). preprint, not peer reviewed; opened 2026-11-18 all records citing this
- Himmelstein DS, Lizee A, Hessler C, et al.. Systematic integration of biomedical knowledge prioritizes drugs for repurposing. eLife 6:e26726 (2017). paper; opened 2026-11-18 all records citing this
- Huang K, et al., Zitnik M. Zero-shot drug repurposing with geometric deep learning and clinician centered design (TxGNN, medRxiv version 4). medRxiv (PMC11326339) (2024). preprint, not peer reviewed; opened 2026-11-18 all records citing this
Where it sits on the map
History of this record (5)
This record has been changed 5 times, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.
- Kind
- big-open
- Register
- open-question
- Status
- partly written
- Last updated
- 2026-11-19
- Record
- content/questions/what-fraction-of-a-graph-benchmark-score-survives-a-degree-baseline.json