News
In recent years, the influence of machine learning (ML) in the field of protein science has expanded rapidly. From the structural prediction revolution sparked by AlphaFold to the development of various sequence-generation models for protein design, algorithms are now exploring the protein sequence space at an unprecedented pace. However, a fundamental challenge remains: while algorithms can generate hypotheses, the validity of these hypotheses—whether a protein’s function is advantageous or not—must still be validated through experimental data. This is particularly true in enzyme engineering and functional protein optimization, where ML models place even higher demands on training data: the datasets must not only be large but also accurately capture the diversity of protein functions. Traditional workflows that rely on cellular expression, protein purification, and kinetic characterization are often time-consuming and have limited throughput, making them a major bottleneck in the iterative cycle of ML model development. Against this backdrop, a study published by Thornton et al. in ACS Synthetic Biology offers a highly representative solution:
2026-03-06
Service Hotline
Enterprise E-mail
Company Address
Room 302, Building 1, Weili Medical Science Park, 69 Jiepu Road, Suzhou Industrial Park
Product Ordering
Technical Support
