Using generative AI in drug discovery, Iambic Therapeutics (formerly Entos) has advanced its IAM1363 drug candidate from program launch to clinical studies in fewer than 24 months — a process that often takes several years. Iambic Therapeutics’ AI drug development milestone relied on an alliance with NVIDIA researchers and engineers and through the use of AI tools, including NeuralPLexer, a generative AI model designed to predict the three-dimensional structure and binding interactions of protein-ligand complexes.
The company has since developed the next generation of the model, dubbed NeuralPLexer2 that boasts higher accuracy and new features for biomolecular structure prediction and drug design.
IAM1363 is a selective, brain-penetrant inhibitor designed to treat HER2-driven cancers. It is designed to target metastatic tumors throughout the body, including the brain, while offering a wider therapeutic index and reduced toxicity over existing therapies.
Surpassing AlphaFold2 benchmarks, which itself set a new standard for protein folding prediction
In a benchmark study featured on the cover of Nature Machine Intelligence, NeuralPLexer outperformed AlphaFold2 (AF2) and a ligand-free baseline model. Specifically, NeuralPLexer achieved an average TM-score of 0.934, outperforming AF2 (0.929) and the baseline model (0.925). Accurate protein-ligand complex prediction is essential for identifying and optimizing drug candidates.
NeuralPLexer also demonstrated strong performance in blind protein-ligand docking (a computational method for predicting the binding poses of a protein and the corresponding binding poses of a ligand without prior knowledge of the binding pocket location), significantly improving accuracy over existing methods. The system also effectively recovered binding site structures, highlighting its ability to predict protein structures that change as a result of ligand binding.
Iambic Therapeutics AI drug development strategy also prioritizes speed as well as accuracy
Iambic highlights NeuralPLexer2’s prediction speed as another core advantage, stating it has a roughly 50-fold acceleration relative to AlphaFold2. The company says that NeuralPLexer2 is already accelerating compound design in their internal drug discovery programs, including IAM1363.
Iambic Therapeutics worked with NVIDIA over multiple years to develop the NeuralPLexer technology.
Iambic’s AI-driven platform also includes OrbNet, a suite of tools it claims are capable of modeling drug candidate chemical properties 1000 times faster than traditional techniques.
Iambic part of a wave of AI-driven drug discovery companies
A growing number of companies are demonstrating the potential of AI to accelerate traditional drug discovery timelines. For instance, Insilico Medicine’s AI platforms, PandaOmics and Chemistry42, enabled the discovery of a novel drug target for idiopathic pulmonary fibrosis (IPF) and the development of a potent inhibitor within 18 months. Similarly, Lantern Pharma’s AI-powered platform is on track to advance drugs from concept to phase 3 trials for a fraction of the conventional cost.
Pharma giants such as Genentech and Amgen have also aligned themselves with NVIDIA to sharpen their focus on AI-based drug discovery and development.
Filed Under: clinical trials, Drug Discovery, machine learning and AI, Oncology, Uncategorized
