Our machine learning platform, EVA, discovers and produces enhanced proteins using a unique combination of machine learning, robotic automation and advanced gene synthesis.
Proteins can be optimised through a process called directed evolution where variants are selected based on superior properties. Traditionally this process relies on random mutations where the link between protein sequence and function is not fully understood.
We use machine learning models that better understand the relationship between DNA sequence and a protein function. The models make predictions based on data produced in our London-based lab and generate new DNA sequence libraries for us to test rapidly in an ongoing evolutionary cycle. The selection data is fed continuously into the models making them increasingly accurate at predicting functional sequences.
Our machine learning models need to make intelligent predictions about DNA sequences and protein function. To do this, they need high quality data that contains accurate information about protein characteristics. We use robotic automation to generate high quality datasets that make EVA even smarter.
Using our proprietary gene synthesis technology, we can accurately manufacture trillions of unique DNA sequences (termed libraries). Unlike traditional synthesis methods, our technology does not involve either long-sequence hybridisation or polymerase chain reaction (PCR) steps. As a consequence, our libraries do not suffer from either diversity loss (through sub-pool amplification & polymerase-specific biasing) or low fidelity (due to miss-annealing between DNA strands). We have a patent pending on its novel multiplexed gene synthesis technology.