Why depression needs a different drug discovery strategy

Drug discovery for major depressive disorder remains challenging. Patients with similar symptoms can have different underlying biology, making it difficult to identify and validate relevant therapeutic targets.

The Multi-Modal Assessment and Phenotyping in Depression (MAP-D) programme aims to address this by studying depression at a biological, clinical and behavioural level. Managed by the Foundation for the National Institutes of Health (FNIH), the programme will build a large, longitudinal dataset to investigate differences between patients.

More than 300 participants will initially be enrolled in a three-year pilot, with the potential to expand to more than 2,500. Over time, researchers aim to identify biologically meaningful subgroups of depression and validate tools that could support drug development and treatment selection.

Looking beneath the symptoms

Unlike diseases that can be classified using established molecular or pathological markers, depression continues to be diagnosed predominantly through clinical symptoms and patient-reported outcomes.

Steve Hoffmann, Senior Vice President and Chief Preclinical Officer at FNIH, explained that this can result in patients with different biological drivers being grouped under the same diagnosis.

“Depression is considered a heterogeneous syndrome, meaning it can look different from person to person and is still diagnosed through subjective clinical symptoms and patient-reported outcomes. Depression involves complex interactions among brain circuits, neurotransmitters, hormones, immune responses, genetics and environmental and social influences.”

Depression is considered a heterogeneous syndrome, meaning it can look different from person to person and is still diagnosed through subjective clinical symptoms and patient-reported outcomes.

This heterogeneity can also make it difficult to identify and validate therapeutic targets.

“Patients with similar symptoms are often grouped under the same diagnosis, even though their depression may arise from very different biological causes. This diagnostic heterogeneity, high placebo rates and a lack of clearly defined mechanistic targets impede the development of targeted treatments,” Hoffmann explained.

MAP-D will investigate whether different types of data can help identify biologically meaningful patient groups.

The programme will collect genomic, neuroimaging, biological and clinical data alongside patient-reported outcomes. Measures will include electroencephalography (EEG), cognitive assessments, sleep monitoring and digital measures such as speech and facial recognition.

Researchers will analyse these data over time to identify biological signals that may not be apparent from individual measures alone.

Finding biomarkers that are useful for drug discovery

According to Hoffmann, candidate biomarkers must demonstrate analytical validity, reproducibility across sites and populations, biological relevance and clear clinical utility.

“Historically, many promising biomarkers in psychiatry have failed to replicate in larger studies and depression’s underlying heterogeneity has made it difficult to distinguish meaningful biological signals from noise,” said Hoffmann.

MAP-D is therefore initially focused on developing, testing and validating the tools and measures used to characterise depression, including determining which measurements are reliable, meaningful and clinically relevant.

Historically, many promising biomarkers in psychiatry have failed to replicate in larger studies and depression’s underlying heterogeneity has made it difficult to distinguish meaningful biological signals from noise.

Standardising how data are collected will be important if associations identified within the dataset are to progress beyond exploratory findings. Longitudinal validation, harmonised protocols and testing across different populations will also be required.

For drug discovery, the aim is to establish whether particular biological signatures are linked to therapeutically relevant mechanisms, rather than simply associated with a depression diagnosis.

“Those biological signatures can then serve as starting points for identifying molecular pathways, neural circuits and disease processes that represent promising therapeutic targets,” Hoffmann explained. “By linking biological signals to real-world outcomes, researchers can distinguish biomarkers that are merely associated with depression from those that may reflect causal disease mechanisms.”

Connecting multimodal data with AI

The challenge is identifying and interpreting relationships across these different types of data.

MAP-D plans to use artificial intelligence and machine learning to analyse information including genomics, imaging, biological assays, digital phenotyping and clinical outcomes.

Rather than treating these datasets independently, computational models could identify patient clusters or biological relationships that are difficult to detect using conventional analyses.

“AI and machine learning methods can help identify patterns, relationships and patient clusters that would be impossible to detect through conventional analyses. This will allow researchers to uncover biologically meaningful subgroups of depression that may share common mechanisms regardless of whether they appear clinically similar or different,” Hoffmann said.

AI and machine learning methods can help identify patterns, relationships and patient clusters that would be impossible to detect through conventional analyses.

Combining different data types could help researchers identify biological patterns relevant to drug discovery. For example, genetic associations could be examined alongside changes in brain circuitry, molecular measurements and behavioural data collected over time. As more data are collected, researchers could use these patterns to investigate the mechanisms underlying different forms of depression.

“By identifying convergent biological patterns across these domains, as MAP-D expands and evolves longitudinally, AI may also help uncover novel therapeutic targets, predict treatment response and generate actionable biomarkers for future clinical trials,” Hoffmann added.

These computational associations will still require biological and clinical validation. MAP-D is intended to provide a standardised dataset in which candidate biomarkers and patient subgroups can be tested and validated.

The consequences of trial-and-error treatment

The ability to predict treatment response could also help address one of the longstanding difficulties in depression care – finding an effective treatment for an individual patient.

Marty Parrish shared his experience of living with depression and the difficulties he faced in finding an effective treatment. His first major episode occurred at 17, when he was unable to get out of bed for three days. Further episodes followed, sometimes lasting weeks or months. At the time, depression was rarely discussed as an illness and Parrish initially viewed his symptoms as a personal and spiritual failure.

He did not immediately seek professional treatment and instead used alcohol to temporarily relieve his symptoms. This led to a prolonged struggle with alcoholism. In 1987, Parrish sought medical help and was later diagnosed with anxious depression. His first antidepressant was effective, although it took almost six weeks to work. He stopped taking it after three months because of the side effects.

Over the following decades, Parrish tried dozens of medications. Only two were effective for the periods in which he could tolerate their side effects. He also continued to use alcohol until undergoing intensive outpatient treatment and becoming sober in 2011. His depression, however, continued.

In 2014, following several months of severe depression, Parrish underwent transcranial magnetic stimulation (TMS), a non-invasive treatment that uses magnetic pulses to stimulate areas of the brain involved in mood regulation. He has since required two booster treatments but says TMS made a lasting difference.

Parrish also described the time involved in establishing whether an antidepressant is effective.

“You have to try a medication for 4-6 weeks to even determine if it might be working and then, if it’s determined that it isn’t, you have to start all over again. There are negative side effects even for the best-known depression medications that result in patients dropping them before they have time to work,” Parrish said.

You have to try a medication for 4-6 weeks to even determine if it might be working and then, if it’s determined that it isn’t, you have to start all over again.

He also described the effect that repeated trial-and-error treatment can have on patients.

“We are already depressed when we go seeking treatment. We are suffering and in pain, along with our loved ones and supporting friends. This trial-and-error process and lack of relief is very discouraging and causes many people to just quit trying.”

MAP-D will investigate whether biological differences between patients could help predict treatment response and guide treatment selection.

“I am hopeful that MAP-D will result in answers and advances that are sorely needed. I’d like to remind others that the work that MAP-D is doing can mean the difference between life and death for many people,” Parrish said.

What MAP-D could mean for target discovery

MAP-D will not immediately identify new depression targets or establish biological subtypes. Its first phase will focus on developing and validating the measurements needed to investigate these questions.

Depression research has identified numerous potential biomarkers, but reproducibility and translation remain challenging. MAP-D’s standardised longitudinal dataset could help determine whether biological signals are consistent across populations and linked to disease mechanisms and treatment outcomes.

“This creates a much stronger foundation for target discovery, helping industry and academic investigators prioritise mechanisms with the greatest likelihood of clinical relevance,” Hoffmann concluded.

The longer-term ambition is to create a shared research resource that supports target validation, patient stratification and the development of therapies for biologically defined groups rather than a broad diagnostic population.