GLP-1 Modality Bakeoff — Why "The Next GLP-1" Is an Indication Question, Not a Chemistry Question

Semaglutide (Ozempic/Wegovy, Novo Nordisk) hit $42 billion in 2024. Tirzepatide (Mounjaro/Zepbound, Eli Lilly) hit $36 billion. Retatrutide (Lilly Phase III) and orforglipron (Lilly Phase III oral) are queued behind. Every pharma in the world is asking the same question: what's the next GLP-1? The answer depends on what you're treating — not what you'd like to put in a vial.

The receptor pattern decides everything

The GLP-1 family has three receptors, and what changes between the approved and emerging drugs is which of those receptors get hit:

Retatrutide is the only triple agonist. GCGR is the third receptor — and that is the receptor that is the liability for T2D and the therapeutic for MASH. Glucagon raises blood sugar (bad for T2D) and burns hepatic fat (good for MASH). GLP1R is broadly expressed across pancreas, brainstem, and adipose. GIPR is adipose-dominant. But GCGR is hepatocyte-dominant — that one data point from the single-cell atlas is what separates the MASH answer from the T2D answer.

The 4-phase receptor bakeoff

Why "which GLP-1 should I develop?" doesn't have one answer

The receptor pattern isn't arbitrary drug design — it's downstream of the biology of the target indication. Glucagon (GCGR's endogenous ligand) has two clinically relevant effects:

1. Hepatic glucose output — raises blood glucose. For T2D patients already managing hyperglycemia, GCGR agonism is counterproductive. This is why triple-agonist programs had a harder time in T2D: you're adding a glucose-raising receptor to a glucose-lowering drug.
2. Hepatic lipid oxidation and fat burning — GCGR activation reduces hepatic fat accumulation by increasing beta-oxidation. For MASH, where the primary problem is hepatic steatosis, GCGR agonism is directly therapeutic.

The atlas data makes this explicit: GCGR expression is concentrated in hepatocytes, the cell type whose dysfunction underlies MASH. That's the biological justification for why retatrutide-class programs are prioritizing MASH in Phase III while pulling back on T2D positioning.

What BioMate does in 4 phases

Phase A — Receptor atlas profile. Pulls GLP1R, GIPR, and GCGR tissue expression from Tabula Sapiens single-cell data. The atlas data is what makes the indication question answerable from biology rather than clinical intuition.

Phase B — Modality scoring. Scores 6 candidate classes across 6 axes: receptor coverage match, weight-loss potency, metabolic-correction potency (HbA1c, lipids, hepatic fat), safety profile, delivery format, and IP whitespace. Each candidate gets a numeric score per axis.

Phase C — Indication weighting. Applies indication-specific multipliers to the scoring matrix. For MASH, GCGR engagement gets a +10 bonus (the therapeutic mechanism). For T2D, GCGR engagement gets a −10 penalty (the glucagon counter-regulation liability). The weights are exposed as editable parameters.

Phase D — Top recommendation. Outputs a ranked CSV plus rationale narrative plus clinical-program comparator citations. For MASH: retatrutide-class wins. For T2D: tirzepatide-class wins. For cost-sensitive obesity (Indian subcontinent, LMIC): orforglipron wins on oral delivery economics.

Why this isn't a single-prompt LLM question

Ask any general-purpose LLM "what's the next GLP-1 for MASH?" and you'll get a thoughtful paragraph mentioning semaglutide, tirzepatide, and retatrutide without a definitive ranking. It won't give you a ranked answer because ranking requires three things that general LLMs can't provide:

• The receptor expression heatmap from a single-cell atlas (not in pre-training; requires a live query)
• The mechanistic weighting — GCGR good for MASH, bad for T2D — which is a published consensus but is blended in pre-training with conflicting prior literature
• The IP whitespace check (Lilly already has retatrutide in Phase III; developing another triple agonist has a freedom-to-operate question)

BioMate combines all three deterministically. Atlas query is live data. Scoring matrix is canonical with literature citations. Indication weights are exposed as overridable parameters — so a program team can adjust the GCGR weight if they have proprietary data showing different tissue behavior.

"Run the same workflow with indication=t2d and the top recommendation flips from retatrutide-class to tirzepatide-class. Same pipeline. Different weights."

The indication-indexing insight

The practical takeaway for any new GLP-1 program is this: the receptor pattern you choose should be downstream of the indication you're targeting, not upstream of it. Programs that start with "we want to make a triple agonist" and then look for indications are working backward. Programs that start with "our target indication is MASH; what receptor pattern does that require?" get the right drug faster.

BioMate makes the indexing explicit: the same pipeline, the same scoring matrix, the same data. Change one parameter and get a different drug. The answer isn't "which GLP-1 is best" — it's "which GLP-1 is best for which indication." BioMate returns six answers indexed by six indications in one run.

Try it yourself

Compare GLP-1 modalities for MASH: semaglutide vs tirzepatide vs retatrutide

→ biomate.ai · 30 seconds · 4 phases · retatrutide-class returns at the top with a GCGR-engagement rationale.

Switch to indication=T2D → tirzepatide-class wins. Switch to indication=obesity, market=India → orforglipron wins on oral delivery economics. The pipeline propagates the indication change through all four phases.

Further reading: Davies et al. 2021 — Tirzepatide vs semaglutide, SURPASS-2 (NEJM); Jastreboff et al. 2023 — Retatrutide Phase 2 for obesity (Lancet); Tabula Sapiens Consortium (Chan Zuckerberg Biohub); NIDDK — MASH (NASH) biology.