Towards High Quality Single-cell Experiments: Approaches, Applications and Performance

Abstract

Single-cell RNA sequencing has revolutionized the way molecular mechanisms were being studied by allowing the dissection of gene expression at single-cell resolution. The data acquired from scRNA-seq provides great opportunities for scientist to push the limits and go beyond technological boundaries to address biological questions. However, a thoroughly thought experimental design, protocol selection and data analysis strategies are necessary to get the best out of this high potential technology. In this thesis we start with summarizing current methodological and analytical options, and discuss their suitability for a range of research scenarios. We provide information about best practices in every step from separating cells and RNA library preparation to data generation, normalization and analysis. Next, we try to address a biological phenomenon using scRNA-seq. We demonstrate how a correctly designed scRNA-seq experiment and analysis is able to capture in details the process of dermal fibroblast aging. Observing the data produced by different scRNA-seq protocols, their important differences and the challenge to analyse them together, raised the question of their suitability specially in cell atlas projects. Hence, in a big multi-center systematic study we compared 13 commonly used single-cell and single-nucleus RNA-seq protocols using a highly heterogeneous reference sample resource. We pointed at their accuracy, application across distinct cell properties, potential to disclose tissue heterogeneity, reproducibility and integratability with other methods; features in which should be considered when defining guidelines and standards for international consortia, such as the Human Cell Atlas project. Finally, we propose an approach to elevate the data from poor-performing protocols to the quality of the best data coming from best-performing ones using variational autoencoders and vector arithmetic.