insitro has completed what it says is the first population-scale human genetics study of brown adipose tissue, using artificial intelligence to unlock genetic discovery in a tissue that has historically been difficult to analyse at scale.

The work combines AI-derived imaging phenotypes with genome-wide association studies to identify genes linked to brown adipose tissue biology and metabolic health. The company also reported preclinical validation of a prioritised target, BAT-01, showing significant weight loss in obese mouse models without loss of lean mass.

Brown adipose tissue, often referred to as brown fat, has been increasingly associated with improved metabolic outcomes, including glucose control and lipid metabolism. However, its diffuse anatomy and functional variability have made it difficult to measure in large human cohorts, limiting genetic insight.

insitro addressed this challenge by developing a machine learning-based brown adipose tissue phenotype using MRI data from nearly 70,000 participants in the UK Biobank. The approach relied on computer vision to quantify fat signal differences between anatomical regions known to differ in brown fat content.

Daphne Koller, founder and chief executive officer of insitro, said: “To power human genetics, you need tens of thousands of people – but brown fat measurements have historically required specialised approaches such as PET scans, which are hard to acquire at scale. AI enables these measurements to be derived from broadly available MRIs, unlocking a first-ever analysis of human genetics for BAT.”

According to the company, the phenotype demonstrated biological specificity, including seasonal variation consistent with known brown fat behaviour, with stronger signals observed in winter months. This pattern was not seen in general adiposity measures, supporting the relevance of the approach.

Using this phenotype, insitro conducted a genome-wide association study and identified multiple genetic loci associated with brown adipose tissue. Several of the genes highlighted had not been detected in previous obesity-focused genetic studies, reflecting the novelty of the tissue-specific phenotype.

The company then used its CellML platform to functionally screen genetically supported targets in primary human adipocytes. High-content imaging, transcriptomic analysis and functional assays were applied to assess beige and brown-like adipocyte characteristics, leading to the prioritisation of BAT-01 for in vivo testing.

In diet-induced obese mice, knockdown of BAT-01 using fat-targeting siRNA resulted in a 15 percent reduction in body weight over four weeks. Fat mass was reduced by 25 percent, while lean mass was preserved. The intervention did not affect food intake, suggesting a mechanism distinct from centrally acting appetite suppressants.

David Lloyd, senior vice president of metabolic disease and translational pharmacology at insitro, said: “Starting with scalable human phenotypes and genetic support allows us to move into functional validation with far more confidence and conviction.”

He added: “These preclinical results point to BAT-linked targets that promote fat loss and cardiometabolic health through selective peripheral targeting while avoiding appetite suppression.”

Molecular analysis of adipose tissue showed increased Ucp1 expression and reduced Leptin expression in white fat depots, findings consistent with the induction of a beige-like adipocyte state.

The research was presented at the Keystone Symposia on Obesity Therapeutics. Insitro said it is continuing to evaluate additional brown adipose tissue-linked genes identified through the study, with the aim of building a differentiated pipeline of obesity and cardiometabolic disease targets.