Search

International audience; Plant genomes are characterized by large and complex gene families that often result in similar and partially overlapping functions 1. This genetic redundancy severely hampers current efforts to uncover novel phenotypes, delaying basic genetic research and breeding programs 2. Here, we describe the development and validation of Multi-Knock, a genome-scale CRISPR toolbox that overcomes functional redundancy in Arabidopsis by simultaneously targeting multiple gene-family members, thus identifying genetically hidden components. We computationally designed 59,129 optimal single guide RNAs (sgRNAs) that each target 2 to 10 genes within a family at once. Furthermore, partitioning the library into ten sub-libraries directed towards a different functional group allows flexible and targeted genetic screens. From the 5,635 sgRNAs targeting the plant transportome, we generated over 3,500 independent Arabidopsis lines that allowed us to identify and characterize the first known cytokinin tonoplast-localized transporters in plants. With the ability to overcome functional redundancy in plants at the genome-scale level, the developed strategy can be readily deployed by scientists and breeders for basic research and to expedite breeding efforts.

It is well known that for some tasks, labeled data sets may be hard to gather. Self-training, or pseudo-labeling, tackles the problem of having insufficient training data. In the self-training scheme, the classifier is first trained on a limited, labeled dataset, and after that it is trained on an additional, unlabeled dataset, using its own predictions as labels, provided those predictions are made with high enough confidence. Using credible interval based on MC-dropout as a confidence measure, the proposed method is able to gain substantially better results comparing to several other pseudo-labeling methods and out-performs the former state-of-the-art pseudo-labeling technique by 7\(\%\) on the MNIST dataset. In addition to learning from large and static unlabeled datasets, the suggested approach may be more suitable than others as an online learning method where the classifier keeps getting new unlabeled data. The approach may be also applicable in the recent method of pseudo-gradients for training long sequential neural networks.