An Integrated Computational-Experimental Strategy For the Prediction of Small Molecules as GLP-1R Agonists

This study presents an integrated computational-experimental framework that successfully identified diverse GLP-1R agonist candidates, including the pentapeptide DPDPE, which exhibits full agonism and dual GLP-1R/GIPR activity, thereby establishing a robust strategy for discovering chemotype-diverse therapeutics against flexible GPCR targets.

The Gist

Imagine your body has a very important "thermostat" for hunger and blood sugar. This thermostat is controlled by a specific lock on the surface of your cells called the GLP-1 receptor. When the right key turns in this lock, it tells your body to feel full, stop eating, and manage sugar levels.

For years, doctors have used "keys" made of long, complex chains (peptides) like Semaglutide (the drug Ozempic) to open this lock. They work great, but they have two big problems:

1. They are expensive and hard to make into pills (most must be injected).
2. Sometimes, when people stop taking them, they gain the weight back.

Scientists wanted to find a small, simple key (a tiny molecule) that could do the same job as the big, complex ones, but be easier to swallow and cheaper to make. The problem? The lock is weird. It's flexible, changes shape, and is very picky. Trying to find a small key that fits is like trying to guess the shape of a moving target in the dark.

The "Super-Scanner" Strategy

Instead of guessing, the scientists in this paper built a digital "Super-Scanner."

Think of it like a massive treasure hunt. They had a library with over one million different digital keys (chemical compounds). To find the right one, they didn't just use one map; they used five different maps at the same time:

• Map 1 & 2: Looking for keys that look similar to known good keys.
• Map 3: Checking if the electrical charge of the key matches the lock.
• Map 4: Building a 3D "ghost" of the perfect key and seeing which real keys fit inside it.
• Map 5: Physically simulating how the key fits into the lock on a computer.

By combining all these maps, they created a "Consensus Strategy." If a key only showed up on one map, they ignored it. But if a key appeared on multiple maps, it was a strong candidate. This prevented them from getting tricked by false alarms.

The Three Winners

Out of the million keys, the Super-Scanner picked three special candidates to test in the real world:

1. GQB47810: A tiny, non-peptide molecule. It looked promising on the computer, but when tested, it was a bit weak. It turned the lock, but not very hard.
2. Neuromedin C: A small peptide (a short chain of amino acids). It fit the lock well, but again, it wasn't strong enough to be a great medicine.
3. DPDPE (The Star): This was a tiny peptide (only 5 building blocks long) that scientists already knew from other research (it's related to pain relief).

The Big Surprise

When they tested DPDPE, something amazing happened.

• It worked perfectly: It turned the GLP-1 lock just as effectively as the big, expensive drugs, telling the body to stop eating and manage sugar.
• The "Double Agent" Effect: But here's the twist. DPDPE didn't just open the GLP-1 lock; it also opened a second lock called GIPR.
  • Analogy: Imagine you have a house with two doors: the Front Door (GLP-1) and the Back Door (GIP). Most drugs only open the Front Door. DPDPE is special because it has a master key that opens both doors at the same time.
  • Why does this matter? Opening both doors might be even better for weight loss and blood sugar control than opening just one. It's like having a "dual-action" therapy.

The Catch (and the Future)

There is one small issue: DPDPE isn't as strong as the current drugs. You need a higher dose to get the same effect (like turning a key with more force). Also, because it's related to painkillers, it might affect the brain's "pleasure" centers if the dose is too high.

The Takeaway:
This paper proves that you don't need a giant, complex key to open the hunger lock. A tiny, simple key can work just as well if you find the right shape. By using a smart computer strategy that combines many different ways of looking at the problem, the scientists found a "dual-action" candidate (DPDPE) that could lead to a new generation of weight-loss drugs—perhaps even a pill that works better than the injections we use today.

In short: They built a digital sieve to filter a million keys, found a tiny, double-opening key, and proved that small things can have a massive impact on how we treat obesity.

Technical Summary

1. Problem Statement

The Glucagon-like peptide-1 receptor (GLP-1R) is a Class B1 G-protein-coupled receptor (GPCR) critical for glucose homeostasis and appetite regulation. While peptide-based agonists (e.g., semaglutide) are clinically effective, they face limitations including the need for parenteral administration, gastrointestinal side effects, and recent concerns regarding weight regain upon cessation. Consequently, there is a high demand for orally available, non-peptidic small-molecule agonists.

However, discovering these small molecules is exceptionally difficult due to:

• Structural Complexity: GLP-1R possesses large extracellular domains (ECD) and an extended orthosteric binding site.
• Conformational Plasticity: The receptor exhibits multiple active-state conformations, making single-conformation docking prone to bias.
• Methodological Bias: Conventional virtual screening (VS) relying on a single method (either ligand-based or structure-based) often fails to recover diverse chemotypes, limiting scaffold diversity and hit rates.

2. Methodology

The authors developed an integrated computational-experimental pipeline that combines Ligand-Based Virtual Screening (LBVS) and Structure-Based Virtual Screening (SBVS) under a consensus-driven selection strategy.

A. Computational Workflow

• Library Preparation: Over one million compounds were curated from DrugBank, FooDB, Biosynth, MolPort, and Enamine. Structures were standardized, protonated at physiological pH, and converted to 3D conformers.
• Structure-Based Virtual Screening (SBVS):
  • Target Structures: Two active-state GLP-1R crystal structures were used: PDB 6ORV (complexed with TT-OAD2) and PDB 6XOX (complexed with orforglipron).
  • Docking: Performed using Lead Finder (LF) within the MetaScreener platform. A consensus approach retained only compounds showing reproducible binding poses across both receptor conformations.
  • Pharmacophore Modeling: Structure-based pharmacophores were generated using LigandScout based on key interactions in the active sites. Screening required a fit score ≥ 0.70.
• Ligand-Based Virtual Screening (LBVS):
  • Reference Ligands: A diverse panel of known agonists (TT-OAD2, orforglipron, danuglipron, CHU-128, etc.) was used.
  • Descriptor & Fingerprint Analysis: Used QikProp for physicochemical descriptors and FPscreener for multi-fingerprint similarity (ECFP, topological, path-based).
  • Electrostatic Similarity: Utilized OpenEye's ROCS (shape) and EON (electrostatics) to align candidates with reference ligands.
  • Ligand-Based Pharmacophores: Pairwise shared-feature pharmacophores were generated to capture conserved interaction motifs across different chemotypes.
• Consensus Strategy: Hits were prioritized only if they appeared in the intersection of multiple screening methods (e.g., docking + pharmacophore + fingerprint similarity) to minimize false positives and methodological bias.

B. Experimental Validation

• Cell Lines: HEK293T cells (for cAMP and $\beta$-arrestin assays) and CHO cells (for receptor selectivity).
• Functional Assays:
  • cAMP Accumulation: Measured via TR-FRET (LANCE Ultra kit) to assess Gs-mediated signaling.
  • $\beta$-Arrestin Recruitment: Measured via BRET to assess biased signaling.
  • Selectivity: Tested against GIPR and GCGR to determine dual/triple agonism.

3. Key Contributions

1. Robust Consensus Framework: Demonstrated that a multi-method, multi-conformation consensus approach significantly improves hit identification for structurally complex Class B1 GPCRs compared to single-method workflows.
2. Discovery of Diverse Chemotypes: Successfully identified three chemically distinct agonist candidates:
   • GQB47810: A non-peptidic small molecule.
   • Neuromedin C: A peptide.
   • DPDPE (2,5-Pen-enkephalin): A small pentapeptide.
3. Novel Mechanism Identification: Revealed that a small peptide (DPDPE) can act as a full agonist with efficacy comparable to large endogenous peptides, challenging the paradigm that bulky scaffolds are required for Class B1 GPCR activation.
4. Dual Agonist Discovery: Identified DPDPE as a dual GLP-1R/GIPR agonist, a profile previously associated only with large peptide therapeutics like tirzepatide.

4. Key Results

• Hit Rate: From a prioritized list of 58 compounds, three showed reproducible agonistic activity in vitro.
• GQB47810:
  • Prediction: Identified via multi-fingerprint consensus; docked into the upper/central TM2-TM3 bundle, interacting with residues A200, L201, M204, Y220, and W297.
  • Result: Showed low potency and modest efficacy in vitro.
• Neuromedin C:
  • Prediction: Identified via docking; engaged the canonical peptide-binding groove with extensive ECD and TM contacts.
  • Result: Showed low potency; lacked key interactions with residues W297 and R380 required for high efficacy.
• DPDPE (2,5-Pen-enkephalin):
  • Prediction: Identified via structure-based pharmacophore screening.
  • In Vitro Efficacy: Exhibited full agonism at GLP-1R, reaching maximal efficacy ($E_{max}$) comparable to endogenous GLP-1 and danuglipron, though with lower potency ($EC_{50}$ in the 1 $\mu$M range).
  • Selectivity: Demonstrated dual agonism at GLP-1R and GIPR with similar efficacy profiles. It showed no activity at GCGR.
  • Signaling: Primarily Gs-mediated (cAMP) with no significant $\beta$-arrestin recruitment bias observed, similar to non-peptidic controls like CHU-128.
  • Mechanism: Despite being a small peptide, it engaged hydrophobic residues (Y145, Y148) in the TM1-2 region, distinct from the deep orthosteric binding of large peptides but sufficient for activation.

5. Significance and Implications

• Therapeutic Potential: DPDPE represents a promising "first-in-class" lead for triple-acting ligands (GLP-1R/GIPR/$\delta$-opioid receptor). Since $\delta$-opioid receptor (DOR) activation improves glucose homeostasis and regulates hedonic eating, a small molecule/peptide hybrid targeting all three could offer superior obesity management compared to current single-receptor agonists.
• Paradigm Shift: The study challenges the notion that Class B1 GPCRs require large peptide scaffolds for activation. It proves that small peptides can retain full functional efficacy if key pharmacophoric motifs are preserved.
• Methodological Impact: The proposed integrated workflow serves as a transferable blueprint for drug discovery against other flexible GPCR targets, effectively reducing chemotype bias and improving the robustness of virtual screening campaigns.
• Future Directions: The authors suggest optimizing DPDPE derivatives to improve potency while managing the risk of DOR-mediated sedation, aiming for a balanced therapeutic profile that targets both homeostatic and hedonic pathways of food intake.