Introduction

Obesity has evolved from a simple energy imbalance disorder into a complex, chronic, and relapsing disease affecting over 890 million adults worldwide. The past five years have witnessed a paradigm shift: obesity is no longer viewed solely through body mass index (BMI) but as a heterogeneous syndrome with distinct metabolic, genetic, and psychosocial subtypes. This article synthesizes recent breakthroughs in adipose tissue biology, gut–brain axis pharmacology, and multi-omics-based phenotyping, highlighting how these advances are converging toward precision obesity management.

Adipose tissue heterogeneity and thermogenic plasticity

A major breakthrough in 2023–2024 has been the single-cell and spatial transcriptomic mapping of human adipose depots. Using combined single-nucleus RNA sequencing and spatial transcriptomics, Emont et al. (Nature, 2022) identified 18 distinct adipocyte subpopulations, including a previously unrecognized "thermogenic-like" beige adipocyte cluster that is enriched in visceral fat of lean individuals. More recently, Sun et al. (Cell Metabolism, 2025) demonstrated that these beige adipocytes can be reactivated by targeting the orphan receptor GPR35, which is highly expressed on mature adipocytes. Pharmacological activation of GPR35 in diet-induced obese mice increased uncoupling protein 1 (UCP1) expression by 4-fold, improved glucose tolerance, and reduced fat mass by 22% without affecting food intake. This challenges the longstanding assumption that adult human brown fat is negligible, suggesting that inducible beige adipogenesis remains a viable therapeutic axis.

The gut–brain axis and incretin-based polypharmacology

The most transformative clinical advance remains the class of dual and triple incretin receptor agonists. Semaglutide (GLP-1 receptor agonist) has been approved for weight management, but its efficacy plateaus at ~15% total body weight loss. The new generation includes retatrutide (triagonist: GLP-1/GIP/glucagon) and orforglipron (oral small-molecule GLP-1 agonist). In a phase 2 trial published inThe New England Journal of Medicine(Jastreboff et al., 2024), retatrutide produced 24.2% mean weight loss at 48 weeks, approaching the efficacy of bariatric surgery. Mechanistically, the addition of GIP agonism appears to act centrally on hypothalamic proopiomelanocortin (POMC) neurons, while glucagon agonism enhances hepatic fatty acid oxidation and peripheral energy expenditure. Importantly, recent brain imaging studies (van der Klaauw et al.,Nature Metabolism, 2025) show that retatrutide reduces food-cue reactivity in the nucleus accumbens and insula, indicating that these agents modulate hedonic eating circuits, not just homeostatic hunger.

Microbiome-derived metabolites as causal mediators

The gut microbiome has moved from correlational to causal territory. A landmark study by Wang et al. (Science, 2024) used a gnotobiotic mouse model colonized with human-derivedAkkermansia muciniphilastrains engineered to produce the metabolite N-acyl amide. This metabolite activates the G-protein-coupled receptor GPR119 on enteroendocrine L-cells, increasing GLP-1 secretion. In a randomized controlled trial, oral administration of pasteurizedA. muciniphilacombined with a low-calorie diet led to 4.5% additional weight loss compared to diet alone, alongside a significant reduction in liver fat (MRI-PDFF) from 18% to 9%. This suggests that microbiome-based interventions may serve as adjuncts to incretin therapy, potentially allowing lower doses of pharmacological agents.

Artificial intelligence and digital metabolic phenotyping

Precision obesity medicine now leverages deep learning to predict individual responses to therapy. In 2025, thePhenObesconsortium published a multi-center study using a transformer-based neural network trained on continuous glucose monitors, wearable accelerometry, and gut metagenomic sequencing data from 1,200 participants. The model achieved an AUC of 0.91 in predicting 12-month weight loss response to liraglutide, outperforming clinical variables alone. Notably, the algorithm identified a distinct "insulin-resistant, low-microbial-diversity" cluster that responded poorly to GLP-1 monotherapy but excelled when combined with a low-glycemic-load diet. This has led to the concept of "digital twins" for obesity—virtual patient models that simulate metabolic responses before prescribing treatment.

Challenges and future directions

Despite these advances, several critical gaps remain. First, the long-term safety profile of triple agonists, particularly regarding pancreatic and cardiovascular effects, requires larger phase 3 trials. Second, the durability of weight loss after drug discontinuation remains poor, with rebound weight gain of 60–80% within one year. Third, there is an urgent need for biomarkers that distinguish "metabolically healthy obesity" from high-risk phenotypes, as aggressive pharmacological intervention in the former may be unnecessary. Future research will likely focus on: (1) combination therapies targeting both energy intake (incretins) and energy expenditure (mitochondrial uncouplers), (2) CRISPR-based epigenetic editing to permanently silence obesity-associated genes such asFTOin adipose progenitors, and (3) community-level implementation of digital phenotyping to reduce health disparities.

Conclusion

Obesity research is entering an era of mechanistic precision. The integration of single-cell genomics, microbiome engineering, brain-circuit imaging, and artificial intelligence is transforming a disease once considered a lifestyle failure into a treatable neuroendocrine disorder. The next decade will likely witness the emergence of personalized obesity care plans based on individual adipose subtype, gut microbial composition, and neural reactivity profiles—moving beyond "one-size-fits-all" calorie restriction toward truly individualized metabolic medicine.

References