Citrini 借 Anthropic Claude 药物发现实验发布生命科学多头清单，点名 Charles River、Revvity、Tecan 为实验验证环节受益标的

Life sciences companies stand to gain from a surge in artificial intelligence fueled drug discovery, says Citrini Research after testing Anthropic PBC’s Claude Science.

Charles River Laboratories International Inc. and Revvity Inc. are among the independent investment research firm’s top life sciences picks that could benefit from AI’s current progress in drug discovery. Citrini’s latest call comes after it used the Anthropic tool to create a new drug candidate, a process that revealed AI can speed up the time for research and spur demand for experimental validation.

“As these ‘virtual biologists’ become more widely adopted, demand should rise for any company that sends them their first real-world invoice,” Citrini analysts wrote in a note on Tuesday.

The use of AI is becoming a routine part of the race to develop powerful new medicines across numerous diseases, with drugmakers increasingly using AI to identify and design promising new molecules as well as salvage medicines that might otherwise have been abandoned. Stocks of AI-linked biotechs have surged over the past year on bets that computers can quickly tackle scientific problems that have historically taken researchers years to solve.

“In a world where discovery is compressed and demand for experimental validation expands, the contract research organization becomes akin to cloud infrastructure for biology,” the researchers added, referring to firms that drug companies hire to manage and run clinical studies. “Cells, assays, reagents, and patient samples become like compute.”

Read: Anthropic Crispr-Like Findings Draw Caution From Scientists

Charles River, for instance, is one of the largest providers of monkeys used for drug testing, an important area of medical research that’s suffered a shortage in recent years.

“To the degree this trade keeps going, and the degree that people focus on bottlenecks, there’s just about nothing more bottlenecked in biotech than monkeys.”

For Revvity, which makes instruments and software for research, its exposure to the experimental work surrounding AI drug design is a positive, according to Citrini.

Revvity makes many of the “picks and shovels” for research and development, with the analysts dubbing the firm as “the compute of biotech.”

Citrini also called out lab automation firm Tecan Group AG among its top stock picks.

Tecan “addresses the different (and potentially larger) constraint of how to convert a large number of proposed experiments into repeatable laboratory work without increasing manual coordination at the same rate.”

The life sciences sector has seen a reversal of the pandemic boom, given a difficult funding environment, and years of investors finding more exciting places for their money. While a 19-member index of life sciences stocks has steadily recovered from the post Covid lull and is up over 18% so far this year, the gauge is still down more than 10% from its pandemic-era peak.

“A neglected sector with the chance for an AI narrative, it had the makings of our favorite type of setup — a secular story at cyclical lows,” Citrini said.

Claude Experiment

Prior to the bullish call, a group of Citrini analysts partnered with Anthropic to test the role of artificial intelligence in drug discovery. In a paper titled Citrini Therapeutics: We Wanted to Use Claude to Create a New Drug. Here’s What Happened, Citrini set up in Anthropic’s offices, using its local network and an internal, general-purpose AI research model to create a new drug candidate.

Citrini, with the support of Anthropic’s Life Sciences team, decided to create a molecule that could work as a scorpion antivenom. It took just days for Claude Science to produce about 180 molecule designs that fit the roughly 35-page prompt the researchers had provided.

“AI will drastically compress the time it takes to turn an idea into an experiment,” Citrini researchers wrote in the report. “That means democratization.”

The designed proteins were sent to Twist Bioscience Corp. for further synthesis, with 17 out of 211 designs achieving the strict criterion of binding to at least one target toxin out of the five toxins they were tested against.

“This was a good result,” the researchers found. “Going in, our expectations were quite low. We didn’t really expect to design anything that actually bound.”

The experiment hasn’t yet tested whether the designed proteins actually stop the toxin.

While researchers viewed the Claude experiment as generally successful in its ability to compress “a year worth of work into a couple days,” they acknowledged the difficulty for AI when it comes to solving the complexity and unpredictability of how drugs work in the human body.

Still, Citrini argues that AI can help produce better treatments against familiar targets and that the process of making an existing drug more practical can unlock huge value.

“We don’t need AI to cure cancer to have a massive impact,” the researchers wrote. “What matters for the market right now is not where AI will ultimately take medicine, but what it will change over the next few years.”