Evoselect® is our machine learning enzyme engineering technology. It leverages evolutionary information inherent in sequence alignments of a homology family to generate novel sequences with improved properties.
Key Differentiators
EvoSelect® integrates seamlessly into the "Design Phase" of our IsoZym™ platform. Our platform supports the full spectrum of enzyme discovery and development. Each step is executed as a standalone work package or combined to support end-to-end enzyme production. By partnering with us, you gain access to a comprehensive suite of services, ensuring consistency, efficiency, and accelerated timelines. Moreover, Isomerase provides consultancy support, including risk assessment and technoeconomic analysis. Additionally, we offer insights into cost optimisation, risk mitigation, and sustainable innovation to ensure economic viability and operational excellence.
How does the EvoSelect® machine learning enzyme engineering work?
EvoSelect® uses unsupervised learning to identify statistical patterns from sequence alignments of homology families, drawn from proprietary sequence databases. Coevolutionary analysis captures inter-residue relationships (couplings) within a protein, which are used to build a probabilistic pairwise model (i.e., Potts model[2]). Establishing these couplings allows us to incorporate epistasis into mutational fitness prediction, resulting in improved accuracy compared to single-site models. Therefore, a generative probabilistic interpretation is built out of the model, from which we iteratively sample mutations, simulating evolution and generating new sequences with improved fitness.
Example pipeline using EvoSelect®
EvoSelect® streamlines enzyme development from target selection to material supply in just 8 weeks. Target sequence input is followed by ML-driven sequence optimisation, where EvoSelect® iteratively mutates sequences to align with target criteria. Following codon optimisation, 40 enzyme candidates are overexpressed in our IsoChassis® hosts and evaluated for expression and target activity. Afterwards, the top-performing candidate is selected, and utilizing our drop-in high cell density fermentation methods, we produce 1-10 grams of the optimized enzyme. Therefore, our process is designed for royalty-free commercialisation, allowing freedom to operate without downstream licensing fees.
Contact us
Reach out to explore how our machine learning enzyme engineering platform will refine and accelerate your enzyme development, bringing your innovations to market faster.