The Next Generation of GLP-1s: PYY Analogs, AI-Designed Molecules, and Weekly Insulin

PYY3-36 is a gut hormone your small intestine releases after a meal. It tells the brain you are full, and it nudges glucose handling in a useful direction. The catch is that native PYY has a short half-life and hits multiple Y receptors when the interesting biology — appetite, glucose — lives mostly at Y2.

A 2025 paper in Science Translational Medicine describes what amounts to a careful rebuild. The authors ran a variant screen of PYY3-36 to find the amino-acid substitutions that locked in Y2 selectivity, then attached a fatty diacid chain so the molecule sticks around in the bloodstream. The result was a class of long-acting, highly Y2-selective analogs that improved glucose metabolism in diabetic mice.

The more interesting result, for anyone watching the obesity pipeline, came when those analogs were stacked with a long-acting GLP-1 agonist. In diabetic rats the combination lowered blood glucose more than the GLP-1 alone; in a high-fat-diet mouse model, it produced greater body-weight loss than the GLP-1 analog by itself. One of the candidates, PYY1875, has now moved into clinical trials for obesity.

Keep the rating in mind: this is animal data plus an early-stage human program. It is a strong proof of concept that the GLP-1 / PYY combination is more than additive marketing — it is biochemistry. It is not yet evidence that you, specifically, will lose more weight on a stack than on the best monotherapy available today.

The second story is less about a molecule and more about a method. A 2025 Drug Discovery Today review argues that artificial intelligence is now a pivotal tool in anti-obesity drug discovery, with a particular focus on GLP-1 receptor agonists and the broader hunt for anti-obesity peptides. In practice that means models that screen vast chemical libraries in silico, predict which peptides will bind well and survive in the body, and prioritize candidates for the wet lab to actually make.

The honest read on this is mixed. AI is plausibly compressing the timeline between "interesting target" and "molecule worth testing," which is the slow, expensive part of drug development. The same review flags the unresolved issues that anyone selling you an AI-discovered miracle will skip past: data quality, integration with existing pipelines, and the fact that existing anti-obesity drugs still struggle with efficacy ceilings, side effects, weight regain, and cost. AI may help with the first half of that list. It will not, on its own, fix the second half.

What it means for a 40-year-old reader: expect more candidate molecules to enter trials over the next several years, and expect louder marketing. The discipline is the same as ever — wait for the human outcomes data.

The third thread is the least glamorous and arguably the most practical. Once-weekly insulin icodec is designed to replace a daily basal injection in type 2 diabetes — one shot instead of seven. The interesting question for the modern patient is whether icodec still behaves itself if you are already on a GLP-1 agonist or an SGLT2 inhibitor, which a lot of well-managed type 2 patients now are.

A post-hoc analysis of the ONWARDS 1–5 trials, published in Diabetes, Obesity & Metabolism, took that question head-on. At baseline, 21.3% of the 3,765 participants were on a GLP-1 receptor agonist and 36.9% were on an SGLT2 inhibitor. The authors then asked whether icodec performed differently in those subgroups versus daily basal insulin comparators.

The short answer is: not really. Across trials, there were no statistically significant treatment interactions by GLP-1 or SGLT2 subgroup for change in HbA1c, change in body weight, weekly basal insulin dose, or achievement of HbA1c under 7% without clinically significant or severe hypoglycemia — with a couple of body-weight and dosing exceptions in ONWARDS 5. Rates of clinically significant or severe hypoglycemia stayed below one episode per patient-year across the trials except ONWARDS 4, the basal-bolus study.

Translation: in this analysis, weekly icodec held its own against daily basal insulin, and it did so whether or not patients were already running a modern metabolic stack. That is the kind of unsexy data point that quietly changes how clinics treat people.

If you are a 40-year-old optimizing energy, body composition and metabolic health, here is the honest take. The current generation of GLP-1s — and the GLP-1/GIP combinations — remain the most powerful pharmacological tool clinicians have for weight and glycemic control, and they are still the relevant conversation to have with your doctor today. The pipeline behind them is genuinely promising, but it is mostly animal data, early human trials, and review-article extrapolation. Promising is not proven.

What does look like a directional bet rather than a guess: metabolic medicine is moving from single-drug heroics toward layered, longer-acting regimens. PYY analogs as a partner for GLP-1s. AI shortening the discovery cycle. Weekly insulin where insulin is still needed. The likely winners will be patients whose clinicians know how to combine these tools intelligently — not whoever chases the next molecule on a forum.

The metabolic toolkit a 40-year-old man will have access to in five years is going to look meaningfully different from the one he has today. Not because of one miracle peptide, but because the field is finally building a stack worth stacking.