Advanced quantitative analyses of long-read RNA-seq at the bulk and single-cell dimensions

Abstract Long-read sequencing has been demonstrated powerful for investigating gene isoforms, which could have critical biomedical roles, such as in embryogenesis, neuroscience and cancer. In the past decade, the Au lab has been at the forefront of developing novel bioinformatics software for identifying complex gene isoforms in different perspectives and have delivered many new transcriptomics insights at the gene isoform level via the applications to stem cells and cancer. Beyond the development of 8 long-read RNA-seq software, the Au lab also published 4 papers for establishing the corresponding data collection and analysis guidelines. In particular, the Au lab is one of the primary organizers of the US-Europe consortium “Long-read RNA-seq Genome Annotation Assessment Project (LRGASP)” for evaluating the effectiveness and the methodology of long-read RNA-seq. In total, the Au lab as the primary authors has published 14 papers focusing on long-read sequencing with >4,800 citations. Beyond the existing qualitative analyses and basic quantification, the greatly improved yield of long-read sequencing makes advanced quantitative analyses with long-read RNA-seq possible, which will be of wide utility towards studying gene isoforms in depth. Especially, with bulk data, comparative analysis of gene isoform abundance can find the meaningful biomedical differences; with single-cell data, cell clustering at the gene isoform level has potential to find new cell types that have been masked by gene-level analysis in the past. Indeed, the prevalence of fruitful studies of alternative splicing has indicated the great potential of the gene isoform-level analysis in delivering biomedical discovery. The Au lab’s accumulating expertise in the field is well-suited to establish the methodological foundation of this goal in this funding cycle. Our previous studies have shown that varying quantification error exists in gene isoform quantification, which depends on gene isoform structure complexity and data. To tackle this varying error in the advanced quantitative analyses, we will develop a set of bioinformatics methods using read count as direct input to represent gene isoform profile. Our proposed methods will allow discovery of differential expressed isoforms, allele bias at isoforms and isoform switching with statistical significance from bulk long- read RNA-seq; and allow discovery of new cell types with unique gene isoform expression from long-read single-cell RNA-seq. In parallel, we will establish the corresponding guideline for data collection and analysis as we did in the first wave of long-read RNA-seq bioinformatics in the past decade. We will apply these methods to our own data and the publicly available data from multiple consortia for investigating gene isoforms quantitatively at stem cells, neuroscience and other broad biomedical contexts.