CRISPR researcher calls Anthropic enzyme find routine genome mining
Lucas Harrington of Mammoth Biosciences says the search method is decades old and that similar systems have been known since 2008.
A named CRISPR researcher has disputed the framing of Anthropic's claim, reported here on Tuesday, that Claude agents found a previously uncharacterised enzyme system in DNA sequence databases. Lucas Harrington, a co-founder of Mammoth Biosciences who trained in Jennifer Doudna's laboratory, said the method behind the search has been around for decades.
Harrington said similar systems have been known since 2008, according to The Decoder, which reported his comments on Thursday. "The hard part is figuring out what a system actually does," he said, and presenting early results as a major discovery "isn't helpful". The Decoder is the only outlet carrying his remarks, and Anthropic has not responded to them publicly.
What the system does remains the open question. Anthropic said the array-associated reverse transcriptases its agents surfaced pair a repeating DNA pattern with a reverse transcriptase gene, and called that arrangement reminiscent of CRISPR. The company also said its work on the primary function was still ongoing, and no other group has reported reproducing the result.
Feng Zhang of MIT and the Broad Institute, the outside researcher Anthropic cited on Tuesday, called the finding intriguing and worth further investigation without saying it was confirmed. Harrington's objection is narrower and different in kind: not that the result is wrong, but that database mining of this sort is ordinary work.
TechCrunch, writing on Tuesday, took a different detail as the notable one — that Anthropic has not let Claude run unsupervised in its laboratory, and that humans remain in the loop. Anthropic's post reached 748 points and 761 comments on Hacker News by Thursday evening.