Replication audit of published hub genes in osteoarthritis and intervertebral disc degeneration: data and code deposit

Data and code for a prespecified, held-out audit of hub genes reported by bioinformatics studies of osteoarthritis (OA) and intervertebral disc degeneration (IDD). For both diseases the deposit covers two corpora, distinguished by what was available at the freeze: the existing corpora, papers identified by the searches whose full texts were available and analysed before the analysis plan was frozen (exploratory), and the held-out sets, papers identified by the same searches whose full texts were not retrievable at the freeze, which were registered at freezing and retrieved and analysed only afterwards (confirmatory). It contains the frozen analysis plan (SAP v1.0 and Amendment v1.1) with its SHA-256 freeze record; the search strings, dates, counts and screening records for both corpora; two isolated extraction passes with their adjudication and blinded human verification; the coding manual and the extraction tables (paper level, dataset roles, hub genes); the quality-graded test pools with processed expression matrices and per-gene effect sizes; the pair-level replication database with the matched null draws; source data for every figure and table; the analysis code; and the complete deviation log. For readers the deposit also carries English copies of the tables, English translations of the frozen protocol and of the deviation logs, and a Chinese-to-English data dictionary covering every table that retains Chinese headers. In the confirmatory held-out OA set (74 papers, 255 gene pairs) 56.9% of hub genes replicated the published direction against 51.4% expected by chance (one-sided P = 0.044 pair-weighted, 0.007 paper-weighted); the held-out IDD analysis (12 papers, 58 pairs) was descriptive (39.7%); in the IDD existing corpus (225 studies from 231 reports, of which 22 papers and 76 pairs were evaluable) 40.8% of pairs agreed with the published direction. No article full text is redistributed. Data are licensed CC BY 4.0; code in 06_code/ is licensed MIT.