The Challenge
Peptide therapeutics are among the most important medicines available today, yet their manufacture carries significant environmental and financial costs. Current production methods rely on toxic solvents, protected amino acids, and inefficient stepwise processes—generating millions of kilograms of hazardous waste annually. The world's largest peptide manufacturer alone produces over 15.1 million kilograms of hazardous waste each year.
With demand for peptide-based treatments surging dramatically due to GLP-1 therapies like Ozempic and Mounjaro, the pharmaceutical industry faces an urgent challenge. Scaling production using existing methods would multiply an already unsustainable environmental footprint. The sector needs a fundamentally different approach—one that can meet growing demand while eliminating the toxic byproducts that define conventional peptide synthesis.
Our Approach
Our collaboration with Origin Peptides began with predictive modelling to identify which peptides would synthesise successfully using their novel aqueous templated process—the only truly solvent-free peptide synthesis method. We engineered over 150 descriptors per peptide, incorporating chemical properties, sequence patterns, and structural characteristics derived from AlphaFold protein structure predictions. Deploying optimised XGBoost models, we improved experimental planning and resource allocation from the outset.
This foundational work contributed to securing a £6.4 million award from Innovate UK's Sustainable Medicines Manufacturing Innovation Programme (SMIP), in partnership with Origin Peptides, the Centre for Process Innovation, and Queen Mary University London. The programme is now scaling aqueous peptide synthesis to commercial production levels.
New Gradient is leading AI and software development for the programme. We are building machine learning systems for real-time process optimisation—analysing data as peptides are synthesised and automatically adjusting production conditions throughout the manufacturing run. Using model predictive control and frontier neural network architectures, our models must generalise across diverse peptides and operating conditions with limited training data, optimising both initial parameters and continuous adjustments throughout synthesis.
The Outcome
The combination of Origin Peptides' novel aqueous process and our real-time ML optimisation delivers transformative results. The system achieves 6x faster production compared to conventional methods while generating zero toxic waste—addressing both the efficiency and sustainability challenges facing the industry.
The platform includes industry-leading live QA/QC capabilities, with in-line sampling feeding our AI-powered optimisation software to enable continuous improvement throughout production. This closed-loop approach represents a first for peptide manufacturing, allowing the system to learn and adapt in real time rather than relying on post-hoc analysis.
As the programme scales toward commercial deployment, this work positions sustainable peptide manufacturing as a viable alternative to conventional methods. The approach demonstrates how machine learning can enable entirely new manufacturing paradigms—not just optimising existing processes, but facilitating a completely novel process to be executed in the most intelligent and efficient way, at a commercially viable scale, not possible without the ML systems we have generated.