Iambic Therapeutics’ NeuralPLexer consistently outperforms AlphaFold2, according to a new report. The study looked at global protein structure accuracy on both representative structure pairs with large conformational changes and recently determined ligand-binding proteins. Iambic says NeuralPLexer is able to directly predict protein–ligand complex structures solely using protein sequence and ligand molecular graph inputs.
The report was published online this week in Nature Machine Intelligence. This work was a collaboration between scientists at Iambic, Caltech, and NVIDIA. The senior author was NVIDIA’s Animashree Anandkumar.
“Iambic’s NeuralPLexer2 is pushing the boundaries of generative AI in 3D protein prediction, helping to enable new capabilities by accurately representing how structures alter their shape as a result of drug interactions,” said Rory Kelleher, global head of business development for life sciences at NVIDIA.
Iambic has emerged as one of the stars in AI for drug discovery and development. The company announced NeuralPLexer in Sept. 2023 as a “first-in-class generative AI model addressing this grand challenge problem of jointly folding protein-ligand complex structures, and predicting their dynamics in a structure ensemble.” In October, 2023, Iambic had an oversubscribed $100 million series B financing to advance AI-discovered therapeutics.
The company has also released a white paper highlighting the improvements NeuralPLexer2. They say the program has already demonstrated significant improvements in it’s prediction accuracy and the model has been scaled to include most categories of biological structures adding protein-protein complexes, cofactors, post-translational modifications (PTMs), and protein-nucleic acid complexes, and encompassing almost all structures in the Protein Data Bank (PDB).
“NeuralPLexer is allowing us to discover pharmacological patterns for increasingly complex protein targets and target areas and achieve unprecedented selectivity, novel mechanisms of structural engagement, the ability to expand patient populations by adding multiple target mutations as well as identify new mechanisms of action at protein-protein interfaces and at other unspecified sites. We are now generating structures that once took many months and significant investment to generate in just a matter of seconds.”
Iambic Therapeutics uses NeuralPLexer in building its own pipeline. The program was used for the discovery of IAM1363, a selective and brain-penetrant small molecule inhibitor of HER2 wildtype and oncogenic mutant proteins, designed to expand therapeutic index compared to available HER2 inhibitors and to avoid toxicities from off-target inhibition of EGFR, a related receptor tyrosine kinase.
In preclinical studies, the company reports IAM1363 has demonstrated over 1000-fold selectivity for HER2 compared to EGFR. An IND for IAM1363 was recently accepted by FDA and clinical trials are planned to commence in early 2024—a timeline that has the drug candidate moving from program start to clinical studies in under two years.
Iambic says its AI-driven platform was created to address the most challenging design problems in drug discovery, incorporating the most current AI technologies and purpose-built tools from Iambic. The company says the integration of physics principles into the platform’s AI architectures, “Improves data efficiency and allows molecular models to venture widely across the space of possible chemical structures.”
The platform’s algorithms enable identification of new chemical mechanisms for engaging difficult-to-address biological targets, discovery of defined product profiles that optimize therapeutic window, and exploration of the chemical space to discover candidates for development with highly differentiated properties. Through close integration of AI-generated molecular designs with automated experimental execution, Iambic completes design-make-test cycles on a weekly cadence.
