Adaptive imaging cytometry to estimate parameters of gene networks models in systems and synthetic biology

The use of microfluidics in live cell imaging allows the acquisition of dense time-series from individual cells that can be perturbed through computer-controlled changes of growth medium. Systems and synthetic biologists frequently perform gene expression studies that require changes in growth conditions to characterize the stability of switches, the transfer function of a genetic device, or the oscillations of gene networks. It is rarely possible to know a priori at what times the various changes should be made, and the success of the experiment is unknown until all of the image processing is completed well after the completion of the experiment. This results in wasted time and resources, due to the need to repeat the experiment to fine-tune the imaging parameters. To overcome this limitation, we have developed an adaptive imaging platform called GenoSIGHT that processes images as they are recorded, and uses the resulting data to make real-time adjustments to experimental conditions. We have validated this closed-loop control of the experiment using galactose-inducible expression of the yellow fluorescent protein Venus in Saccharomyces cerevisiae. We show that adaptive imaging improves the reproducibility of gene expression data resulting in more accurate estimates of gene network parameters while increasing productivity ten-fold.

将微流控技术（microfluidics）应用于活细胞成像（live cell imaging），可采集单个细胞的高密度时间序列数据，并可通过计算机控制更换培养基实现细胞扰动。系统生物学家与合成生物学家常开展基因表达研究，这类研究往往需要改变培养条件，以表征基因开关的稳定性、遗传装置的传递函数，或基因网络的振荡特性。研究人员往往难以预先（a priori）知晓各类条件变更的最佳时机，且唯有在实验结束后完成全部图像处理（image processing）工作，才能知晓实验是否成功。这导致研究者需重复实验以优化成像参数，进而造成时间与资源的浪费。为克服这一局限，我们开发了一款名为GenoSIGHT的自适应成像平台，该平台可在图像录制过程中实时处理图像，并基于所得数据对实验条件进行实时调整。我们通过酿酒酵母（Saccharomyces cerevisiae）中黄色荧光蛋白Venus（yellow fluorescent protein Venus）的半乳糖诱导表达（galactose-inducible expression），验证了该实验闭环控制（closed-loop control）方案的有效性。研究表明，自适应成像可提升基因表达数据的可重复性（reproducibility），从而更精准地估算基因网络参数，同时将实验效率提升十倍。