Developing psychiatric drugs with precision focus

May 27, 2024

Sarah Anderson joined Drug Discovery News as an assistant editor in 2022. She earned her PhD in chemistry and master’s degree in science journalism from Northwestern University and served as managing editor of “Science Unsealed.”

Psychiatrist and neuroscientist Amit Etkin has long been dismayed by the lack of progress in treating psychiatric disorders amidst a growing mental health epidemic. “One thing that becomes imminently clear when you have any interaction with the psychiatric patient population is just how poorly served they are by the existing therapeutics,” said Etkin, who has treated patients as a clinician and studied psychiatric disorders at a fundamental level. He attributes this lapse in care to a limited range of biological mechanisms in the current pool of psychiatric drugs as well as the absence of diagnostic tools to classify disease subtypes.

Motivated to turn the tide with a precision medicine approach, Etkin transitioned from academic research at Stanford University to start the company Alto Neuroscience in 2019. Drawing on a decade’s worth of data on how individuals with psychiatric disorders respond to various therapies, his team developed a pipeline of clinical-stage drugs targeting key disease pathways that operate differently in unique patient subpopulations. The researchers also harnessed this data to build machine learning models that use a patient’s biomarkers to predict his or her response to a drug from their portfolio. By matching a patient’s biomarker profile to a corresponding drug, Etkin’s team aims to remove the guesswork from psychiatric drug development and more efficiently deliver effective treatments for depression and related diseases.

As clinicians, it's incredibly frustrating to give a patient a treatment where we have no idea whether it will work. There is zero predictive value, apart from that the patient or perhaps a family member responded to a particular drug before. We don't know whether it actually works until months later, and this cycle can go on for a long time. For example, patients with depression can easily go through six months of different treatments before they find something that works. That should be frustrating to everybody; people should not accept that as the status quo.

The root of this problem is how these drugs have been developed, which relies on the assumption that a diagnosis such as depression is a meaningful way to categorize patients. In reality, any psychiatric diagnosis is extremely diverse biologically and clinically, so there’s a subpopulation that responds well, but many other patients who don’t. The problem is that we don't know which is which.

Clinical trials often fail because the study populations are diverse in an uncontrolled and unmeasured way. As a field, we learn nothing from these failures or, more shockingly, from the successes. Even if a drug works, we don't really understand why or for whom. This begets yet more trial and error in drug development and deployment of drugs in the clinic.

We can put together a mechanism of disease with a mechanism of intervention much more logically if we measure things about a patient. Consider fever as an example: Fever is a symptom driven by many possible conditions, much like depression is a description of a set of symptoms that can have many different drivers. With a fever, we would perform a number of tests to diagnose the underlying cause in order to intervene with the right treatment. In depression, that has not been the case. The logic is very simple and translatable: if we make biologically meaningful measurements, we can cluster groups of patients within a diagnosis and then align a drug with each patient profile.

Each brain is unique, but there are commonalities between people that allow us to subgroup them biologically. The brain can't be biopsied, so we can't readily access it at a molecular level. However, I would argue that what we need to understand about the brain is not what happens in one tiny part of the brain or in one cell type, but rather how the brain encodes information and engages in the kinds of tasks that someone performs day to day. We can learn a lot about people’s brain function from their cognitive abilities, their capacity to multitask and remember things, their decision making based on rewards or punishments, and the way their circadian rhythm is regulated. Aspects of cognition, emotion, and sleep are really core, and from an engineering perspective, we can feasibly manipulate these brain circuits with specific drugs and measure the response. We don’t anticipate that the effect of the drug will be limited to cognition, emotion, or sleep; rather, we expect that by pulling on these different leverage points in the right people, we will be able to change their overall clinical picture.

ALTO-100 works in part by manipulating brain-derived neurotrophic factor (BDNF) and BDNF-dependent signaling. This growth factor is involved in brain plasticity and is at the center of our understanding of cognition and mood and the function of corresponding brain regions such as the hippocampus. We hypothesized that people with mood disorders such as depression who also have disrupted brain plasticity influencing cognition could benefit from ALTO-100. In a series of well-controlled studies, we found that ALTO-100 works better for people with a specific cognitive profile. We identified the biomarker to select those patients and independently replicated it in a separate group of patients, so we're confident that we’ve found a robust and reliable signal.

Each brain is unique, but there are commonalities between people that allow us to subgroup them biologically. - Amit Etkin, Alto Neuroscience

In our phase 2a clinical trial, we tested ALTO-100 in people with and without the biomarker and compared clinician-rated measures of depression. We found that the difference in the magnitude of the effect between those two cohorts was approximately double, which is a significant enrichment of the response. Importantly, we also found that it didn't matter if we administered the drug by itself or in combination with a conventional antidepressant. This result tells us that testing the drug as a monotherapy or as an adjunctive therapy is far less important than the people we test it in. We now understand for whom a BDNF-driven drug is the right one. We’re very excited about these findings and have launched a phase 2b study where we will test the drug in a much larger population and look at the role of the biomarker. I think that it will be a really important first outcome for precision psychiatry.

We haven't specified the particular biomarker that we're using, but we measure a broad range of biomarkers. When developing our machine learning models, we aim to identify biomarkers that maximize predictive power and minimize patient burden. By relying on biomarker measures that are easily deployed and scaled, we have sharpened our focus to understand which data we need to collect and the simplest and most powerful ways to define a patient. We analyze brain activity and connectivity with electroencephalography (EEG), a form of brain imaging that can be used in the clinic and in the home. We also conduct objective, performance-based tests of cognition and emotion on a computer and measure sleep and circadian rhythms using a wearable device. Even when we go into a study with a certain hypothesis, we collect the full suite of biomarkers because that's important for building our platform. By comparing information between different populations and drugs, the process catalyzes itself to become better, faster, and more predictive.

We’re interested in a broad range of diseases. The disorders that we focus on are related to these core vertical areas of cognition, emotion, and sleep, which includes schizophrenia, bipolar disorder, and post-traumatic stress disorder. Even outside of traditional psychiatry, we see the same kind of attributes, such as in nonmotor symptoms of Parkinson's disease. A large segment of those patients have cognitive and mood problems and look functionally like a psychiatric population although they have a neurological disorder. We don't discriminate between these boundaries; we think of what we can manipulate and what we can measure.

As we see success over the next few years, we’ll be able to branch out to other diseases with the same drugs and biomarkers and quickly understand for whom a drug is the right one biologically. We want to advance the ability to find the right people for a particular drug and also to understand why that population responds better to facilitate the development of the next generation of drugs.

Reflecting on where we are today with these trials, biomarkers, and drugs in hand, it's incredibly fulfilling. There's an inflection happening in the field where the expectation will be precision and real clinical impact. We have to get past the point where a tiny advantage over placebo is good enough, and I'm excited to see Alto Neuroscience lead that charge.

This interview has been condensed and edited for clarity.

Sarah Anderson joined Drug Discovery News as an assistant editor in 2022. She earned her PhD in chemistry and master’s degree in science journalism from Northwestern University and served as managing editor of “Science Unsealed.”

July/August 2023 Issue