MAUD 1.0

MAUD 1.0 (Multi-parametric AI for Unbiased Design) is Menten AI's next-generation computational platform for the de novo design of cyclic peptide therapeutics. By combining generative AI, physics-based modeling, and quantum simulations, MAUD 1.0 addresses a critical gap in drug discovery: up to 80% of drug targets remain beyond the reach of conventional small molecules and biologics, yet peptide macrocycles — ranging from 800 to 1800 Da and 6 to 15 amino acids — occupy precisely the size and property space needed to engage these challenging targets.

Unlike traditional machine learning approaches that depend on large training datasets, MAUD 1.0 leverages physics-informed reinforcement learning to enable efficient molecular design with atomic precision. The platform is intended for drug discovery teams seeking to develop oral peptides, cell-permeable macrocycles, protein-protein interaction (PPI) modulators, and multi-specific peptide binders against targets previously considered undruggable.

Core Platform Workflow

1. Target Assessment: Using only a target protein structure as input, MAUD 1.0 analyzes protein dynamics and structural features to identify optimal binding sites — including previously unrecognized surfaces — thereby expanding the range of druggable targets.
2. Molecular Engine: The physics-based generative AI engine takes the target protein model as input and performs iterative, multi-parameter design cycles. Multiple drug-like properties are optimized simultaneously, enabling the discovery of potent, cell-permeable, and orally bioavailable cyclic peptides for challenging targets.
3. Hit/Lead Optimization: Top hits are refined using MAUD 1.0's optimization engine, which focuses on beneficial amino acid substitutions — both natural and unnatural — to improve single or multiple properties simultaneously, including potency, stability, and permeability. This fully in silico approach accelerates multi-parameter optimization without requiring experimental structural data.

Demonstrated Drug-Like Properties

• Nanomolar binding affinity (under 100 nM)
• Oral bioavailability (greater than 18% F)
• Cell permeability (Log Pe greater than -6)
• High target selectivity
• Favourable drug metabolism and pharmacokinetics (DMPK) profiles

Platform Applications

• Oral Peptides: Design cyclic peptide inhibitors against extracellular or intracellular targets with oral bioavailability incorporated as a design parameter.
• Cell-Permeable Peptides: Apply Menten AI's proprietary platform to design cyclic peptides against intracellular targets, with a key focus on disrupting protein-protein interactions.
• Molecular Glues and Bi-specifics: Functionalize and conjugate peptides to generate multi-specific peptide binders with diverse therapeutic applications.
• Computational Optimization: Perform in silico structure-based optimization with multi-parameter enrichment of drug-like properties, reducing reliance on experimental cycles.

Key Advantages Over Existing Modalities

• Peptide macrocycles combine properties unavailable in either small molecules or biologics, including cell permeability, oral delivery potential, high selectivity, low toxicity, PPI binding capability, and low immunogenicity.
• Physics-informed reinforcement learning eliminates the need for large experimental training datasets.
• Multi-parameter simultaneous optimization reduces the iterative burden of traditional medicinal chemistry campaigns.
• The platform can engage previously unrecognized binding surfaces, broadening the druggable target space.

Menten AI's MAUD 1.0 platform is supported by an active scientific publication record, with peer-reviewed and preprint contributions covering heuristic energy-based cyclic peptide design, conformational autoencoders for macrocyclic backbone sampling, and quantum chemistry integration via RosettaQM, among other areas. The platform is available for partnership engagements targeting a broad range of therapeutic indications.