Creating new functional proteins, such as remediation enzymes, has traditionally been an inefficient and costly ‘trial-and-error’ process. Biotech companies painstakingly make random mutations of existing proteins, test them, and repeat this cycle until success.

AI ‘de-novo’ protein design promised a breakthrough by generating entirely new proteins from scratch. However, it severely falls short due to a lack of comprehensive data, resulting in poor accuracies that make it unfit for industrial use.