everypixel © Seventyfour @ stock.adobe.com
A robotic cloud-based laboratory platform is able to bring together lab instruments and processes with Internet-of-Things technologies to automate research and development, accelerating scientific discovery. By taking advantage of this concept, researchers can remotely operate infrastructure through web interfaces and use standard communication protocols to automate end-to-end laboratory processes in order to improve reproducibility, standardize data collection, cut down on human error, and increase efficiency.
In the specific case of biochemical design and modeling, for instance, it is a field that relies on trial-and-error experimentation, thus imprecise measurement makes it difficult to explore complex systems with reproducible results. In response, researchers began to outsource high-throughput experiments to automated labs, while mainly continuing to use heuristics to design experiments and individual syntheses. Machine Learning can help guide the overall laboratory process by digesting unstructured data such as written lab reports or patents, as well as structured data from sensors or cloud-based measurements.
Chemical retrosynthesis is a perfect example of how machine learning can automate complex processes that previously required human experts. In retrosynthesis, the chemist works in reverse, repeatedly breaking the target compound down into smaller units until it consists solely of primary chemical precursors. Accordingly, discovering this path of synthesis is essentially equivalent to the search problem or the line of strategic thinking involved with games like Go or chess. By using deep neural networks as a guide, the system finds paths from source to target compounds in the vast space of possible chemical reactions. Compared to their human counterparts, these systems can find shorter routes using fewer chromatography steps, while generating higher yields and discovering new methods of developing viable unpatented synthesis paths.
Beyond the Lab: Automating Agricultural Engineering
In agriculture, the development of automated platforms and software for biology research could be a resourceful tool when attempting to engineer solutions for crop production issues. From seed to table, research that collaborates with growers, harvesters, food processors, and the agriculture and food market industries, could provide the necessary support for the development of new technologies that will increase profitability as well as environmental sustainability. This comprehensive approach would include a sensitive fully-connected network of real-time wireless Yield Monitoring, fruit and Livestock Monitoring, remote and contact sensing of soils and plants, automated and targeted pest management systems, vegetable and fruit post-harvest handling, smart and/or robotic agricultural machinery, on-machinery crop-sensing and control systems, and finally, a sensor-based field operations management system.
New data-driven methods will accelerate discovery while helping geographically dispersed teams design and create a cloud-based open data exchange for large-scale collaborative robotic experiments. A universal chemical synthesis machine could be the future of chemical research and production. If given a target compound, the machine would be able to calculate the steps required for synthesis and then carry out the complete reaction using a supply of chemical precursors —essentially additive manufacturing machines for organic compounds that develop their own design process. In the future, these machines could also integrate synthetic biology in order to create more complex compounds such as proteins and metabolites. Using machine learning, scientists can begin to focus their efforts on deciding what to make and why, while leaving the question of how to make certain compounds up to the machine.