Medical LLMs may show real-world potential in identifying individuals with major depressive disorder using WhatsApp voice note recordings

Some models were over 91% accurate for female participants, though this dropped to 78% accuracy for male participants

2026-01-21

(Press-News.org) A new medical large language model (LLM) achieved over 91 percent accuracy in identifying female participants diagnosed with major depressive disorder after analyzing a short WhatsApp audio recording where participants described their week, according to a study published January 21, 2026 in the open-access journal PLOS Mental Health by Victor H. O. Otani, from Santa Casa de São Paulo School of Medical Sciences and Infinity Doctors Inc., Brazil, and colleagues.

Major depressive disorder is a mental health condition that affects over 280 million people globally, and early detection can be critical for timely treatment. Here, Otani and colleagues used machine learning models to classify individuals with and without major depressive disorder based on WhatsApp voice messages.

The authors used two 2 datasets for this study, a dataset to train their LLMs (with 7 different sub-models used) and then a dataset to test their LLMs. The training dataset consisted of 86 participants: a group of outpatients (37 women, 8 men) clinically diagnosed with major depressive disorder and a control group of 41 volunteers (30 women, 11 men) with no depression diagnoses. The dataset used to test the trained models consisted of 74 participants: 33 outpatients (17 women, 16 men) diagnosed with major depressive disorder and 41 control group participants (21 women, 20 men) with no depression diagnoses. All participants were provided informed consent and screened to exclude potential confounding factors such as other medical issues. In the training dataset, outpatient speech data was taken from WhatsApp audio recordings sent to their doctor’s offices when they were symptomatic; control group participants chose their own routine WhatsApp audio voice messages to share. For the test dataset, speech data taken from the outpatient group and the control group were the same: recorded WhatsApp messages counting from 1-10, as well as audio messages describing their past week. All audio messages in both datasets were from native Brazilian Portuguese speakers.

The LLMs showed greater accuracy when classifying women compared to men as depressed VS not-depressed, particularly when given the “describe your week” data, with an accuracy rate of 91.9 percent for the highest-performing model. The highest-performing model’s accuracy when classifying male participants was 75 percent for the same “describe your week” audio. (This may potentially be explained by the higher number of women participants in the model training dataset, as well as differences in speech patterns between men and women.) The LLMs showed more similar performance between men and women when given the “count to 10” data, with the highest-performing model 82 percent accurate in women and 78 percent accurate in men.

The authors are hopeful that continued refinement of their models could produce a low-cost and practical way to screen individuals for depression, as well as other potential clinical/research applications.

Senior author Lucas Marques adds: “Our study shows that subtle acoustic patterns in spontaneous WhatsApp voice messages can help identify depressive profiles with surprising accuracy using machine learning. This opens a promising path for low-burden, real-world digital screening tools that respect people’s daily communication habits.”

In your coverage please use this URL to provide access to the freely available article in PLOS Mental Health: https://plos.io/4sInRpy

Citation: Otani VHO, Aguiar FO, Justino TP, Buck HS, Grilo LB, Figueiredo MF, et al. (2026) ML-based detection of depressive profile through voice analysis in WhatsApp™ audio messages of Brazilian Portuguese Speakers. PLOS Ment Health 3(1): e0000357. https://doi.org/10.1371/journal.pmen.0000357

Author Countries: Brazil

Funding: This work was financially supported by Infinity Doctors Inc. for the following authors: RU, FA, DV, TO, LM, and VO. The funder provided financial support for the time and effort of these authors during the development of this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

END

ELSE PRESS RELEASES FROM THIS DATE:

Early translational study supports the role of high-dose inhaled nitric oxide as a potential antimicrobial therapy

2026-01-21

Overuse of antibiotics has accelerated the development of bacterial resistance to conventional drugs, a global health crisis projected to result in more than 10 million deaths annually by 2050. The multidrug-resistant bacterium Pseudomonas aeruginosa accounts for approximately one-fifth of hospital-acquired pneumonia cases and is associated with severe illness and increased mortality. Nitric oxide is a therapeutic gas that researchers at Massachusetts General Hospital, a founding member of the Mass ...

AI can predict preemies’ path, Stanford Medicine-led study shows

2026-01-21

An artificial intelligence-based tool can predict the medical trajectories of individual premature newborns from blood samples collected soon after they are born, a Stanford Medicine-led study has shown. The research, which will publish Jan. 21 in Science Translational Medicine, provides a new understanding of the complexity of premature birth, not as a single problem defined by early arrival but as several distinct conditions. The study is a step toward predicting and preventing complications of prematurity using treatments ...

A wild potato that changed the story of agriculture in the American Southwest

2026-01-21

Starchy residue preserved in ancient stone tools may rewrite the story of crop domestication in the American Southwest, according to new research led by the University of Utah. The Four Corners Potato (Solanum jamesii) has been an important cultural, nutritional and medicinal food staple across the Colorado Plateau for millennia. Despite its long history and contemporary use, the extent to which Indigenous people domesticated S. jamesii remains unknown. Previous genetic research has shown that the tubers were transported and intentionally cultivated far beyond its natural range—two crucial steps toward ...

Cancer’s super-enhancers may set the map for DNA breaks and repair: A key clue to why tumors become aggressive and genetically unstable

2026-01-21

New study shows that cancer damages its own DNA by pushing key genes to work too hard. Researchers found that the most powerful genetic “on switches” in cancer cells, called super-enhancers, drive unusually intense gene activity. That high gear creates stress on the DNA and can cause dangerous breaks. Cancer cells can often repair this damage, but the process is frequently error-prone, the repeated cycle of breaking and repairing can make these regions more prone to accumulating mutations over time. In short, the same mechanisms that help cancer grow quickly may also make its DNA more fragile, helping explain how tumors continue to evolve and, in some cases, ...

Prehistoric tool made from elephant bone is the oldest discovered in Europe

2026-01-21

A remarkable prehistoric hammer made from elephant bone, dating back nearly half a million years ago, has been uncovered in southern England and analysed by archaeologists from UCL and the Natural History Museum, London. It is the oldest elephant bone tool to ever be discovered in Europe and provides an extraordinary glimpse into the ingenuity of the early human ancestors who made it. The research, published in Science Advances, describes the roughly 500,000-year-old tool, and reveals the unexpectedly sophisticated craftsmanship and skill of the species responsible for making it, likely either early neanderthals or another species known as Homo heidelbergensis. ...

Mineralized dental plaque from the Iron Age provides insight into the diet of the Scythians

2026-01-21

Researchers have deciphered the diet of an important nomadic people in Eastern European history. By analyzing dental calculus, they have provided the first direct evidence that the diet of the Scythians included milk from various ruminants and horses. For centuries, the Scythians have been regarded as a nomadic horsemen people who roamed the vast steppes of Eurasia during the Iron Age. This image remains powerful to this day. In recent years, however, scientific research has challenged this simplified narrative. It shows that the so-called “Scythians” were not a uniform group, but consisted of a diverse, multi-ethnic population with different geographical origins. ...

Salty facts: takeaways have more salt than labels claim

2026-01-21

Some of the UK’s most popular takeaway dishes contain more salt than their labels indicate, with some meals containing more than recommended daily guidelines, new research has shown. Scientists found 47% of takeaway foods that were analysed in the survey exceeded their declared salt levels, with curries, pasta and pizza dishes often failing to match what their menus claim. While not all restaurants provided salt levels on their menus, some meals from independent restaurants in Reading contained more than 10g of salt in a single portion. The UK ...

When scientists build nanoscale architecture to solve textile and pharmaceutical industry challenges

2026-01-21

Scientists from the CSIR-Central Salt and Marine Chemicals Research Institute (CSMCRI), Indian Institute of Technology Gandhinagar, the Nanyang Technological University, Singapore, and the S N Bose National Centre for Basic Sciences have collaborated to develop a new class of highly precise filtration membranes. The research, published in the Journal of the American Chemical Society, could significantly reduce energy consumption and enable large-scale water reuse in industry. Everyday industrial processes, like purifying medicines, cleaning textile dyes, and processing food, rely ...

Massive cloud with metallic winds discovered orbiting mystery object

2026-01-21

Sweeping winds of vaporized metals have been found in a massive cloud that dimmed the light of a star for nearly nine months. This discovery, made with the Gemini South telescope in Chile, one half of the International Gemini Observatory, partly funded by the U.S. National Science Foundation and operated by NSF NOIRLab, offers a rare glimpse into the chaotic and dynamic processes still shaping planetary systems long after their formation. In September 2024, a star 3000 light-years away suddenly became 40 times dimmer than usual, and remained so until May 2025. The star, J0705+0612, is similar to our Sun, so its ...

Old diseases return as settlement pushes into the Amazon rainforest

2026-01-21

Old diseases return as settlement pushes into the Amazon rainforest (Santa Barbara, Calif.) — Human activity continues to expand ever further into wild areas, throwing ecology out of balance. But what begins as an environmental issue often evolves into a human problem. Researchers at UC Santa Barbara investigated how changes in land use may be driving the growth in human yellow fever cases in the Amazon basin. Their analysis, published in Biology Letters, reveals that the growing border between forested and urban areas is causing an alarming uptick in cases. “Yellow fever is increasingly infecting humans ...

LAST 30 PRESS RELEASES:

American Meteorological Society announces new executive director

People with “binge-watching addiction” are more likely to be lonely

Wild potato follows a path to domestication in the American Southwest

General climate advocacy ad campaign received more public engagement compared to more-tailored ad campaign promoting sustainable fashion