Fig. 2: Microfluidics and coding/decoding in DropAI.
From: AI-driven high-throughput droplet screening of cell-free gene expression
a DropAI generates massive combinations in droplet reactors with a multi-droplet microfluidic merger. The device loads a pool of carrier droplets into a microchannel at regular intervals. Each carrier droplet meets 3 satellite droplets at downstream junctions, which are all merged under an electrical field at the micro-teeth structure. b Micrographs depict the droplet pairing and merging process. The false coloring depicts carrier droplets (green) and differentiates the satellite droplets (yellow, pink, and blue for satellite drop 1, 2, and 3, respectively). c As every droplet is coded by fluorescent color and intensity, the merged droplets are rendered a 4-color barcode (FluoreCode) indicating the exact combinations. The microfluidic device generates the combinatorial pool at ~300 Hz, and the total coding capacity is 6561 (94). d A stacked micrograph displaying the as-built combinatorial pool. e For decoding, the merged droplets are imaged under 4 fluorescent channels to extract the FluoreCode of each droplet. The micrographs are representative examples from three independent experiments. The droplets are then clustered based on the FluoreCode information. The font colors indicate the fake colors used in the imaging channels. Alexa-488 (green), Alexa-546 (yellow), Alexa-594 (red), Alexa-647 (blue). f A t-distributed stochastic neighbor embedding (T-SNE) projection of 153,302 droplets containing 6527 recognized clusters, ~99.5% of the entire combinatorial space. The colors are used to represent individual clusters. Scale bars: 100 μm.
