Physics, Machine Learning and Robotics united by Automated reasoning.
Our technology relies on a disruptive combination of physics-based and robotics-inspired modeling; machine and deep learned information extracted from Nature and experimental results, all united by a dedicated automated reasoning engine that efficiently and optimally integrates these different sources of information with custom-defined sequence requirements.
Thanks to this unique combination of features, various protein properties can be targeted for optimization, including thermo-resistance, activity, solubility, specificity, selectivy or affinity.
Starting from fundamental principles and universal data, our technology applies on orphan, de novo designed, as well as existing proteins, including enzymes, binders or self-assembling systems
It can contribute to improve protein-based processes in various fields, including Biotechnology, Chemistry, Bioenergy, Health, Environment and Food.
When structural information is available, we rely on the most recent atomic and quantum molecular modeling force-fields, and scoring functions to choose an amino acid sequences that will make your protein real, effective and resistant to the various stresses it may have to sustain during its existence.
Using all the data available in structure and sequence databases, we exploit the most recent generative Machine and Deep learning technology to extract information on sequence, structure and function relationships from Nature and other experimental data. Thanks to this, ill-defined properties such as expressability can also be optimized.
Proteins can be modeled as complex poly-articulated robotic systems with many degrees-of-freedom. We leverage recent and fast robotics-inspired algorithms to efficiently capture the crucial flexibilities of target proteins, from specific loop movements to essential ligand access/exits paths.
Our final sequence design engine relies on state-of-the-art AI automated reasoning technology to integrate all the information provided by Physics, Machine learning and Robotics with customer specific requirements to efficiently produce a library of diverse sequences that can also account for multiple targeted conformational states.
