A European Project For Precision Psychiatry: This Company Is Using AI To Predict Mental Health Disorders

Daniel Gehrlach, Ph.D. is the associate director of biomarkers at HMNC Brain Health, a biopharma company using genetic analyses and AI to predict and treat mental health disorders through individually tailored interventions.

Gehrlach, who recently gave a presentation at MAPS' 2023 conference told Benzinga how AI could change mental health treatment approaches and HMNC’s state-of-the-art projects.

Why Psychiatric Drugs Fail 

Since psychiatry’s inception almost 50 years ago, patients receive diagnoses based on symptoms. Conditions like like low mood and anhedonia -inability to feel pleasure - persist for two weeks and are combined with additional related symptoms, would likely be diagnosed as Major Depressive Disorder (MDD) and treated with one single medicine.

Gehrlach explained that research has shown there's so much heterogeneity that having the same diagnosis for all cases does not make sense. For example, possible causes leading to depressive symptoms span from chronic inflammation to genetic predisposition or diet, exercise and sleep. 

“The argument is that it's actually a collection of many different diseases that we cluster together as depression. But having one single treatment that covers all of them is not working out, and that's what we see in the clinic,” he said. 

SSRIs and related compounds, studies show, hold patient response rates of between 30% and 50%. For those not responding, doctors usually switch to, or add on, another therapeutic, or a cocktail of drugs, after which still many people still do not respond. That’s where a diagnosis like Treatment-Resistant Depression (TRD) comes in.

“So why do they fail? It's because we do not specifically get to the root of the cause of this patient. And that's what we want to change.”

AIML And Genes As Basis 

HMNC’s current main project consists of finding a predisposition for stress through genes. The team uses genetics because it is “a very available and cheap method,” translatable to any lab and has a fast turnaround time. 

It's "a very stable molecule that gives you more than a snapshot,” as compared to brain scans (fMRI), which show the resolution of the brain's activity within a short time period, which may be subject to variation by factors affecting the body at that moment. It is also more expensive.

The research team began looking at genes and then used Machine-Learning (ML) to train a data-set algorithm to identify patients. The trained set was based on an experimental procedure with consenting patients. 

The results reportedly showed if stress is off or not. HMNC emulated this with its gene-based test and expects that by pairing it with the assessed compound, they acted at the mechanism causing stress excess.

More on treatment response prediction through genes, funding and AI applications for psychedelic medicine coming up.

Photo: Benzinga edit with photo by Gorodenkoff on Shutterstock and Yahoo.