Artificial intelligence-assisted automatic and index-based microbial single-cell sorting system for One-Cell-One-Tube.

Abstract

Identification, sorting, and sequencing of individual cells directly from in situ samples have great potential for in-depth analysis of the structure and function of microbiomes. In this work, based on an artificial intelligence (AI)-assisted object detection model for cell phenotype screening and a cross-interface contact method for single-cell exporting, we developed an automatic and index-based system called EasySort AUTO, where individual microbial cells are sorted and then packaged in a microdroplet and automatically exported in a precisely indexed, "One-Cell-One-Tube" manner. The target cell is automatically identified based on an AI-assisted object detection model and then mobilized via an optical tweezer for sorting. Then, a cross-interface contact microfluidic printing method that we developed enables the automated transfer of cells from the chip to the tube, which leads to coupling with subsequent single-cell culture or sequencing. The efficiency of the system for single-cell printing is >93%. The throughput of the system for single-cell printing is ~120 cells/h. Moreover, >80% of single cells of both yeast and Escherichia coli are culturable, suggesting the superior preservation of cell viability during sorting. Finally, AI-assisted object detection supports automated sorting of target cells with high accuracy from mixed yeast samples, which was validated by downstream single-cell proliferation assays. The automation, index maintenance, and vitality preservation of EasySort AUTO suggest its excellent application potential for single-cell sorting.