Engineering Biology by Design
I work on intelligent biodesign automation by reversing engineering biology through bioretrosynthesis and backstepping retro designs.
We work on automating workflows for synthetic biology planner agile designs. Our mission is to foster sustainable industrial biotechnology of bio-based compounds (biomaterials, pharmaceuticals, next-generation chemicals) through the principles of control engineering and machine learning. I envision building artificial intelligence assistants for bionics systems; from biosensors to bio-inspired robotics.
A Vision of Future Agile Biomanufacturing
Rising as a key technology for the new bioeconomy, the bio-based engineering of fine, specialty and commodity compounds has taken the chemical industry by storm. The once future vision of green, sustainable bioplastics, drug manufacturing and renewable energy sources is suddenly becoming the biotechnology of today. In the transition from an emerging technology into a truly biomanufacturing technology, accelerated reductions of the time from concept to development to scale-up are essential. Pioneering proof of concepts took over 10 years in their making, we cannot afford such latencies anymore. Modern manufacturing biofoundries have adopted an agile approach in order to improve their Design-Build-Test-Learn cycle efficiency. Synthetic biology is becoming in that way intimately intertwined with automation and machine learning technologies. Biofactories delivering at full steam face major design challenges in metabolic pathway designs. Modern manufacturing provides agile synbio design solutions that requires of continuous upgrades based on build and test team interactions. Therefore, rather than keeping a fixed set of design recipes, agile synbio design should allow for a flexible strategy.
One of the approaches is based on optimal experimental synbio design, which provides a good trade-off between experimental capabilities and design space. Design factors can cover a wide range of bottom-up elements such as genetic parts like promoters, gene coding sequences, ribosome binding sites, vector copy number, etc., as well as assembly spatial arrangements, chassis strains or experimental conditions. Such designs are generated in an automated way down to the list of genetic parts to be synthetized. Following the cloud biomanufacturing paradigm, the automated build and test stages proceed through robotic platforms allowing full sample traceability so that experimental results are fed back into the learn and design engines. Statistical analytics are then applied in order to infer an ensemble of models for the design factor-response relationships under either mechanistic hypothesis, i.e., kinetic models, or using model-free approaches, i.e., machine learning. This fully automated design workflow complies with the flexibility requirements of agile design and is therefore, expected to become widely adopted in future biomanufacturing systems.