µIVC-Useq: a microfluidic-assisted high-throughput functional screening in tandem with next-generation sequencing and artificial neural network to rapidly characterize RNA molecules

Abstract

The function of an RNA is intimately linked to its structure. Many approaches encompassing X-ray crystallography, NMR, structural probing, or in silico predictions have been developed to establish structural models, sometimes with a precision down to atomic resolution. Yet these models still require experimental validation through the preparation and functional assay of mutants, which can rapidly become time consuming and laborious. Such limitations can be overcome using high-throughput functional screenings that may not only help in validating the model, but also inform on the mutational robustness of a structural element and the extent to which a sequence can be modified without altering RNA function, an important set of information to assist RNA engineering. We introduced the microfluidic-assisted in vitro compartmentalization (µIVC), an efficient and cost-effective screening strategy in which reactions are performed in picoliter droplets at rates of several thousand per second. We later improved µIVC efficiency by using it in tandem with high-throughput sequencing, though a laborious bioinformatic step was still required at the end of the process. In the present work, we further increased the automation level of the pipeline by implementing an artificial neural network enabling unsupervised bioinformatic analysis. We demonstrate the efficiency of this “µIVC-Useq” technology by rapidly identifying a set of sequences readily accepted by a key domain of the light-up RNA aptamer SRB-2. This work not only shed some new light on the way this aptamer can be engineered, but it also allowed us to easily identify new variants with an up to 10-fold improved performance.