Addressing Sustainability Challenges in Peptide Synthesis with Flow Chemistry and Machine Learning

The high cost and the large amount of toxic waste generated during peptide production overshadow the current technology, requiring the reduction of excess reagents and the replacement of the solvents used. Advances have been made to replace N,N-Dimethylformamide with moderate success. By exploring the solvent parameter space, we have selected several mixtures, tested their swelling ability, amino acid solubility, coupling efficiency, and Fmoc-cleaving capacity, and found the Anisole/DMSO (17:3) mixture to be ideal for coupling. By adjusting the flow parameters, racemization was reduced to < 2% in the case of His, and < 1% for Cys. Several mixtures were screened for optimal Fmoc-cleavage, selected to cover the solvent parameter space uniformly. To test the selected solvent mixtures for aspartimide formation, and Fmoc-cleavage efficiency, both Scorpion Toxin II (VKDGYI) and JR10-mer (WFTTLISTIM) challenging sequences were synthesized. Fmoc-cleavage parameter optimization was performed using a machine learning algorithm (Bayesian Optimization) to reduce aspartimide formation and maximize Fmoc-deprotection. With the final parameters obtained, the Aib-ACP (10-mer), the glucagon like peptide 1 (GLP-1, 30-mer) and bovine pancreatic trypsin inhibitor (BPTI, 58-mer) polypeptides were synthesized with high efficiency and synthetic speed (12 min/cycle). The method is ideal for high temperature synthetic approaches and outperforms current state-of-the-art synthesizers.