Ananya Dutta, Bose Institute
Over the last decade, the fast advancement of techniques for sequencing single-cell transcriptomes has been matched by equally spectacular advancements in computational approaches for analyzing such data. The developing algorithm advances revealed progressively intricate features of the underlying biology, from cell type composition to gene regulation to developmental dynamics, as the capacity and precision of the experimental procedures increased. Simultaneously, fast expansion has necessitated a constant re-evaluation of the underlying statistical models, experiment goals, and sheer quantities of data processing handled by these computational tools. This article discusses the computational steps involved in single-cell RNA sequencing (scRNA-seq) analysis. Evaluating the assumptions made by various techniques, and highlight accomplishments, lingering uncertainties, and limits that should be in mind as scRNA-seq becomes a methodology for researching biology.
Transcriptional states give a high-resolution, comprehensive perspective of genome activity, putting them at the heart of functional genomics research. However, each cell’s state varies, reflecting its function’s purpose, history, and stochastic fluctuations. The merging of scRNA-seq data and computational analysis can uncover the transcriptional underlying for diverse cell states. There are varied computational approaches that follow these steps for formulation and mainly involve: i) quantitative statistical modeling, ii) a data representation in narrow size distribution, iii) an estimate of the expression manifold, with the simplest and most frequent approximate being a set of independent transcriptional subpopulations[1].
The key preprocessing stages in single-cell RNA-seq analysis involve estimation of the transcript profusion, correction of barcode sequencing inaccuracies by Alignment, and molecular counting. Cell filtering and quality control are executed to differentiate empty barcodes/droplets, dead cells. The process of doublet scoring helps to classify potential droplets formed due to co-encapsulation or barcode collisions. Next, the cell sizes are evaluated. Finally, the gene variance analysis is performed. This step-by-step process is evaluated by a series of computational tools available online.
The analysis of the scRNA seq involves narrowing the dataset and find cell-cell similarities. The next step involves capturing complicated, curving cell configurations in the expression space followed by the identification of distinct cell subpopulations and the genes that differentiate them. Trees and curves represent the incessant variation in the cell state.
Peter V. Kharchenko from the Department of Biomedical Informatics, Harvard Medical School, Boston, MA, the U.S.A published in Nature Methods, provides us with detailed insight into the evaluation procedure of single-cell RNA sequencing.
Also read: Anaerobic gut fungi make way for novel antibiotic production
References:
- Kharchenko P. V. (2021). The triumphs and limitations of computational methods for scRNA-seq. Nature methods, 10.1038/s41592-021-01171-x. Advance online publication. https://doi.org/10.1038/s41592-021-01171-x
One thought on “A view into the computational techniques for scRNA-seq”