(636b) Data-Rich Experimentation-Enabled Mass-Activated Droplet Sorting for Screening of Enzymatic Reactions

Directed enzyme evolution is a critical requirement for biocatalytic process development. Very often, this requires screening of several thousands of variants, typically done in well plates, to identify the optimal variant that can enable the desired chemical reaction, making the overall screening workflow expensive in both time and resources. Droplet microfluidics-based platforms offer a promising alternate route as they require minimal reaction volume in nanoliter scales for screening of enzymes and can be operated at high throughput. While optical assays are well-developed to analyze and identify the enzyme-containing droplets, Mass Spectroscopy (MS) based techniques offer universal label-free detection with high throughput. However, MS-based techniques are destructive in nature and some droplet volume must be retained to enable downstream sequencing of any discovered hits. Therefore, the parent droplet must be split into two daughter droplets: one for MS analysis and another for sorting, collection, and DNA sequencing. The splitting of droplets into two channels, however, necessitates the need for an algorithm that allows the one-to-one alignment of droplets in both the channels.

This talk will present our recently developed mass-activated droplet sorting platform that combines (i) MS-based product detection , (ii) camera-based droplet collection , (iii) droplet classification algorithm, (iv) an accurate, robust, and fast temporal sequence alignment algorithm, and (v) a web-based tool to handle data transfer among all these processes that need to be executed in parallel in real-time. The sequence alignment algorithm leverages multiple droplet features, along with temporal information of the droplets to accurately align the two sequences from MS and camera in real time and is agnostic to deviations in size, composition, and temporal separation of droplets detected by camera and MS. Overall, the richness of the data captured during the workflow, in combination with the advanced classification and alignment algorithm enables a data-rich approach that allows us to efficiently and rapidly screen (0.5 droplets per second) several thousands of enzymatic variants in this microfluidic platform.