Introducing ArtiDock 2.0: A New Era in AI-Driven Drug Discovery

Enhancing prediction accuracy and speed in structure-based drug discovery

Introducing ArtiDock 2.0: A New Era in AI-Driven Drug Discovery

Enhancing prediction accuracy and speed in structure-based drug discovery

PARTNERSHIP
Announcement

Summary

Receptor.AI announces the launch of ArtiDock 2.0, a major advancement in AI ligand pose prediction delivering unparalleled accuracy and speed. Built on a proprietary lightweight architecture and trained on augmented real-world datasets, ArtiDock 2.0 improves protein-ligand interaction prediction and addresses key challenges such as pose accuracy and steric clash reduction. Now a core component of high-throughput drug discovery workflows, ArtiDock 2.0 will soon be available on the Nvidia BioNeMo cloud platform.

Full Text

We are pleased to announce the launch of ArtiDock 2.0, an updated version of our AI ligand pose prediction model. This version improves accuracy and efficiency in identifying potential binding sites within protein structures.

ArtiDock 2.0 has been developed to achieve high predictive performance at increased speed. By leveraging a proprietary, lightweight model architecture and augmented data from comprehensive real-world datasets, ArtiDock 2.0 ensures high accuracy in protein-ligand interaction predictions without compromising processing speed.

The introduction of ArtiDock 2.0 addresses key challenges in molecular docking, such as the accurate prediction of ligand poses and the reduction of steric clashes, through enhanced algorithms. This balance of precision and efficiency significantly speeds up the docking process, making ArtiDock 2.0 a crucial tool for high-throughput drug discovery applications.

Supporting this, Figure 1 demonstrates the comparative effectiveness of ArtiDock 2.0 versus established docking methods like Vina and Glide across various RMSD thresholds in the PoseBusters v3 benchmark. This chart underlines ArtiDock 2.0's precision in predicting protein-ligand interactions, which is essential for efficient high-throughput drug discovery.

Figure 1. Performance of ArtiDock and classical docking methods on the PoseBusters v3 Benchmark set (N=308) at different RMSD thresholds.

Early deployment of ArtiDock 2.0 in Receptor.AI’s virtual screening pipeline has shown promising results, including significant hit discovery rates and identification of lead-like compounds with in vivo activity.

ArtiDock 2.0 will soon be available via the Nvidia BioNeMo cloud platform, expanding access to its capabilities for biotech and pharmaceutical users.

The release of ArtiDock 2.0 represents a key milestone in our mission to innovate and enhance drug discovery tools. Further updates and validation data will be shared as the platform continues to evolve.

To learn more about ArtiDock 2.0 capabilities, please refer to our preprint disclosing details about the model architecture, training, and tuning.