GLP-1R (Glucagon-like peptide-1 receptor) is a clinically validated target involved in the regulation of glucose homeostasis, insulin secretion, and appetite control. Activation of GLP-1R has proven therapeutic value in the treatment of type 2 diabetes mellitus and obesity, making the discovery of novel small-molecule agonists an important direction in modern drug discovery.

The OTAVA GLP-1R Agonists Library was developed using an AI-guided cheminformatics workflow designed to identify structurally diverse compounds with predicted agonistic activity toward GLP-1R.

The following computational approaches were employed during the library design process.

1. Bioactivity-Driven Dataset Construction

A training dataset was assembled from the ChEMBL database, focusing on compounds experimentally tested against GLP-1R.

The dataset was curated using the following criteria:

• selection of compounds with reported EC50 values

• normalization of activity data

• removal of incomplete or inconsistent records

After preprocessing, the final dataset contained over 1500 unique molecules with experimentally measured agonistic activity toward GLP-1R.

These data served as the foundation for the machine learning model used in subsequent screening steps.

2. Machine Learning Activity Prediction

A LightGBM regression model was trained to predict the agonistic activity of compounds toward GLP-1R.

Each molecule was encoded using a combination of structural descriptors and molecular fingerprints, including:

• physicochemical descriptors
  (molecular weight, LogP, TPSA, hydrogen bond donors and acceptors, rotatable bonds, ring count)

• MACCS structural keys

• Morgan fingerprints (ECFP4)

This descriptor set allowed the model to capture both physicochemical properties and structural patterns associated with GLP-1R agonists.

The trained model demonstrated strong predictive performance and was subsequently used for large-scale virtual screening.

The model demonstrated strong predictive performance:

• R² = 0.736 on the independent test set

• RMSE = 0.98 pEC50 units

• 5-fold cross-validation R² = 0.78

3. AI-Guided Screening of the OTAVA Chemical Stock

The trained machine learning model was applied to screen the OTAVA compound stock containing approximately 300,000 small molecules.

For each compound, the model predicted its pEC50 value, allowing ranking of molecules based on their estimated agonistic potency.

Top-scoring compounds with predicted sub-nanomolar to low-nanomolar activity were selected for further analysis, resulting in a focused set of approximately 4,000 candidate molecules.

4. Structural Diversity Optimization

To ensure structural diversity within the final library, the selected compounds were further processed using Tanimoto similarity clustering based on Morgan fingerprints.

Clustering allowed grouping structurally similar molecules and limiting the number of close analogs within each cluster.

In addition, Murcko scaffold analysis was performed to identify unique core structures and prevent over-representation of single chemotypes.

This diversity filtering step produced a final library of 2,637 compounds, representing:

• 1,259 structural similarity clusters

• 1,225 unique scaffolds

This balance between predicted activity and chemical diversity ensures broad coverage of chemical space relevant to GLP-1R agonism.

The resulting OTAVA GLP-1R Agonists Library contains 2,637 structurally diverse compounds predicted to possess agonistic activity toward GLP-1R.

The library combines:

• AI-guided activity prediction

• large-scale virtual screening

• structural diversity optimization

making it a valuable resource for discovering novel small-molecule GLP-1R modulators for metabolic disease research.

Request Your Library Today! Fill out the form:

 

Need more information regarding this product? Send us an inquiry: Name: * First Last Company: * E-mail: (if you do not receive a response within 24 hours, please check your email spam folder, as it may have been blocked by your spam filter) * Please specify name of the product: * Word Verification: type_submit_reset_7 SubmitReset