Skip to Main Content

What’s a hallucinogen without the hallucinations? Perhaps a potent and fast-acting antidepressant, according to a new study based on virtual drug screening.

Psychedelic drugs are now widely understood to have striking antidepressant effects, so a group of computational biologists sought to determine whether any new, LSD-adjacent molecules might have other medicinal powers. In work published Wednesday in Nature, the scientists used in silico modeling, generating 3D iterations of more than 75 million related molecules that don’t actually exist, but could.


After virtually testing each and every one, they found therapeutic potential in a handful. Further studies in mice showed that these synthetic, near-psychedelic molecules seemed to have distinct antidepressant activity — minus the trademark hallucinations.

“We didn’t know that could happen,” said Brian Shoichet, a scientist at the University of California, San Francisco, who co-led the research. “It was a hope that it could happen.”

There’s a growing desire to dampen the “trippy” nature of psychedelic drugs, broadening their therapeutic potential. Beginning in 2020, a few different research groups have demonstrated that it’s possible to separate the psychedelic experience from psychedelics themselves, said David Olson, a pharmacology researcher at the University of California, Davis, who has accomplished this feat himself. But this is the first time that such drugs have been conceived virtually.


“I think it’s a really impressive tour de force of screening capabilities,” said Charles Nichols, a professor of pharmacology at Louisiana State University who, like Olson, was not involved in the study. “It’s not really limited to psychedelics: This work could demonstrate proof-of-principle for this massive in silico strategy and paradigm.”

The in silico movement for drug discovery is in its nascence, but researchers like Shoichet have advanced the field over the past four years. Many if not most major biopharma companies are dabbling in this approach to drug development, but no compounds derived this way have yet made it into the clinic.

The existing arsenal of psychedelic drugs, which include synthetic LSD and the mushroom-derived psilocybin, are still in clinical testing. Outside of the colorful and intense hallucinations for which these drugs are known, there aren’t a lot of noted toxicities associated with them.

There are, however, limitations.

Bryan Roth, a psychiatrist and pharmacology researcher at University of North Carolina, Chapel Hill, who co-led the study, noted that “a number of groups of potentially depressed people would not be eligible for treatment with psychedelic drugs.”

“For example,” he said, “if you have a first-degree relative with schizophrenia, it’s recommended you don’t take psychedelics, because it could induce a psychotic-like illness.”

If psychedelics are used in their natural form as a treatment for conditions like depression, PTSD, or anxiety, they need to be administered in a clinic — while patients are proctored for long stretches of time in order to ward away the ill-effects of “bad trips.” This can dramatically increase the cost and complexity of these treatments, Roth said.

“You’re looking at thousands and thousands of dollars that a typical person would have to pay out of pocket,” Roth said. “If you think about rolling out these treatments for the world’s population, there will never be enough therapists for everyone who is depressed.”

The new work began as a “fun idea” discussed over conference coffee by Shoichet and Yale chemist Jon Ellman. They looped in Roth, who received $26.9 million in funding from DARPA aimed at developing better psychiatric medications for depression, anxiety, and substance use disorder without major side effects.

The trio noted that psychedelic drugs like LSD interact with the receptor known as serotonin 5-HT2A. Roth, in particular, has published seminal work detailing the structural biology of the neurotransmitter serotonin and its receptors. Ellman, meanwhile, created a method to synthesize a family of molecules called tetrahydropyridines, which is core to the structure of LSD. Along with Shoichet, they built out the 75 million-strong library of compounds, using a basic molecular infrastructure that could be easily replicable by other labs.

“We used a standard cast of characters — no selenium or anything. Carbons, nitrogens, oxygens, occasionally sulfur,” Shoichet said. “We threw a lot of options at this interesting scaffold, and weird and beautiful flowers came out of that.”

The library started out with millions of potential drug candidates. The researchers then tasked the computer with fitting them against a 3D rendering of the 5-HT2A serotonin receptor.

“It’s like the computer’s trying to figure out a 3D jigsaw puzzle,” Shoichet said.

Ultimately, those millions were whittled down to 17, which were synthesized. Two of these seemed to actually interact with the serotonin receptors, and were tested in mice. And the molecules indeed had antidepressant activity: They were just as effective in mice as fluoxetine, or Prozac, but at a dose that was 40 times lower. And despite the small quantity that was administered, the antidepressant effect of the in silico-derived molecule lasted several weeks.

But what’s a happy mouse? Or a sad mouse? And what does a hallucinating mouse look like?

It’s been observed over decades of psychedelic study that mice, when exposed to hallucinogens, twitch their head in a distinctive way. It correlates well to the human experience of hallucination, Shoichet said.

In one experiment, Shoichet said, mice were put in a stress situation by being suspended by their tails for a few minutes. Untreated mice had some “fight” in them — they struggled to free themselves from the bindings. Mice that were bred to display depressive symptoms, however, stayed slack. But when given the in silico-generated molecules, the depressed mice fought back — struggling in response to the stress test. They did not, however, twitch their heads in the manner characteristic of a mouse hallucination.

While encouraging, Shoichet said, the efficacy still wasn’t anywhere near what researchers wanted. So the team made small modifications to the structure, to ensure that it fit more snugly with the serotonin 5-HT2A receptor. Nichols did point out that he found it surprising that researchers found only two active molecules out of the many millions.

“The simulation methods are imperfect: It’s still riddled with errors, but it gets it right enough,” Shoichet said. “There are lots of false negatives, lots of old shoes when you cast your net.”

The key with in silico modeling, Shoichet said, is to cycle back and forth between prediction and testing — and then optimizing whatever molecules ultimately pass muster. The molecule that the team ultimately patented may have been swimming somewhere in the initial cast of 75 million, but it wasn’t picked up by the virtual testing. And the final molecule in the paper doesn’t actually resemble LSD much at all, Shoichet said. The work appearing in Nature was actually completed two years ago — and a great deal has changed since.

“We have made tremendous progress since then in getting molecules that have better drug-like properties,” Roth said.

The scientists are now trying to make the molecules better, and more selective for serotonin. They’re also trying to carve away unnecessary portions that could have off-target effects or toxicities. In short, they’ve moved away from the computational approach and toward run-of-the-mill drug discovery.

A couple startups have spun out of this approach. Shoichet has launched a company called Deep Apple, which focuses on expanding virtual screening libraries for drug discovery. And Roth has a company focused specifically on these serotonin-targeted drugs — called Onsero.

UCSF, Yale, and UNC-Chapel Hill have a patent on the particular molecules for depression that have come out of this in silico analysis, but the overall library is meant to be available to the public.

As promising as this approach may be, there is some controversy over whether a psychedelic drug can be effective without the psychedelic experience.

“I think people will read this paper and not be fully convinced by it,” Shoichet said. “There are people who will continue to believe that you can’t separate the really psychedelic effects of the molecules from their more therapeutic effects.”

People who take psychedelic drugs often describe the effects as the most profound experience of their lives.The intensity of the hallucinations, and the insights gleaned from them, may play a major role in the antidepressant effect of the drug, said Nichols, the Louisiana State University professor.

In fact, there’s a correlation – though not proven causation – between the subjective intensity of a trip and the ultimate therapeutic effect, he said. But that might just mean that enough of the drug has been administered: Nichols, who is researching psychedelic-based anti-inflammatory drugs, believes it’s possible the antidepressant effects can be purely biological. The LSD molecule is extremely complex, and researchers are still unsure which aspect of its structure creates the antidepressant effect, and which causes the hallucinations.

“We’re able to see antidepressant effects from psychedelics in rats and even fruit flies,” Nichols said. “I don’t imagine fruit flies have existential epiphanies. I think it’s possible for psychedelics to have purely biological antidepressant effects — and can be experienced even without the hallucinations.”

Get your daily dose of health and medicine every weekday with STAT’s free newsletter Morning Rounds. Sign up here.

Create a display name to comment

This name will appear with your comment

There was an error saving your display name. Please check and try again.