Author: Shunfu Mao

Publisher:

Published: 2020

Total Pages: 80

ISBN-13:

DOWNLOAD EBOOK

High throughput sequencing of RNA (RNA-seq) has become a staple in modern molecular biology, with a wide range of applications including RNA transcripts assembly, variants detection, and gene expression estimation for downstream cellular analysis. RNA-seq data is therefore able to provide us with unprecedented insights into cellular organisms. However, they have also introduced a new set of computational challenges because of the nature of the sequenced RNA transcripts and an ever increasing number of RNA-seq experiments. For instance, the RNA transcripts have different expression levels, making the sequenced reads potentially unable to fully cover some lowly expressed gene regions. In addition, the RNA transcripts also share many repetitive patterns, making it ambiguous to determine the regions where some RNA-seq reads are actually sampled. Moreover, there are still many laborious procedures in the RNA-seq data analysis, making it difficult to keep pace with the constantly produced large amounts of RNA-seq data. There is an urgent need for better computational methods that are able to analyze the RNA-seq data more accurately and efficiently. Motivated by this, in the thesis, we have presented novel computational solutions for three computational problems for RNA-seq data analysis: Firstly, we have developed RefShannon - a new genome-guided RNA transcripts (transcriptome) assembly software. RefShannon reconstructs RNA transcripts, based on the alignments of RNA-seq reads onto a reference genome. It exploits the pair-end linking information of RNA-seq reads, and the varying expressions of RNA transcripts, in enabling an accurate reconstruction of the transcripts. Experiments demonstrate RefShannon has superior assembly performance over the state-of-art genome-guided assembly tools. Next, we have developed abSNP - a new RNA-seq SNP calling software. AbSNP detects SNPs in expressed gene regions, based on the alignments of RNA-seq reads onto a reference transcriptome. It exploits the mapping quality scores of RNA-seq reads, and the varying expressions of different genes. AbSNP is a cost-effective method as it requires no additional DNA-seq. It is also able to call SNPs with significantly improved sensitivity in repetitive gene regions, while other RNA-seq SNP callers are unable to make any calls in such regions. Finally, we have developed CellMeSH - a new web server and API package for automatic cell-type identification in single-cell RNA-seq (scRNA-seq) analysis. CellMeSH predicts cell types, based on a set of marker genes as query input. CellMeSH builds its database in a scalable and easy-to-update way using prior literature, and adopts a novel probabilistic method to better query the database. Through a variety of experiments on human and mouse scRNA-seq datasets, CellMeSH has demonstrated richer gene and cell-type information in its database, robust query method, and an overall superior annotation performance.

Download or read book Computational Problems for RNA-seq Data Analysis written by Shunfu Mao and published by . This book was released on 2020 with total page 80 pages. Available in PDF, EPUB and Kindle. Book excerpt: High throughput sequencing of RNA (RNA-seq) has become a staple in modern molecular biology, with a wide range of applications including RNA transcripts assembly, variants detection, and gene expression estimation for downstream cellular analysis. RNA-seq data is therefore able to provide us with unprecedented insights into cellular organisms. However, they have also introduced a new set of computational challenges because of the nature of the sequenced RNA transcripts and an ever increasing number of RNA-seq experiments. For instance, the RNA transcripts have different expression levels, making the sequenced reads potentially unable to fully cover some lowly expressed gene regions. In addition, the RNA transcripts also share many repetitive patterns, making it ambiguous to determine the regions where some RNA-seq reads are actually sampled. Moreover, there are still many laborious procedures in the RNA-seq data analysis, making it difficult to keep pace with the constantly produced large amounts of RNA-seq data. There is an urgent need for better computational methods that are able to analyze the RNA-seq data more accurately and efficiently. Motivated by this, in the thesis, we have presented novel computational solutions for three computational problems for RNA-seq data analysis: Firstly, we have developed RefShannon - a new genome-guided RNA transcripts (transcriptome) assembly software. RefShannon reconstructs RNA transcripts, based on the alignments of RNA-seq reads onto a reference genome. It exploits the pair-end linking information of RNA-seq reads, and the varying expressions of RNA transcripts, in enabling an accurate reconstruction of the transcripts. Experiments demonstrate RefShannon has superior assembly performance over the state-of-art genome-guided assembly tools. Next, we have developed abSNP - a new RNA-seq SNP calling software. AbSNP detects SNPs in expressed gene regions, based on the alignments of RNA-seq reads onto a reference transcriptome. It exploits the mapping quality scores of RNA-seq reads, and the varying expressions of different genes. AbSNP is a cost-effective method as it requires no additional DNA-seq. It is also able to call SNPs with significantly improved sensitivity in repetitive gene regions, while other RNA-seq SNP callers are unable to make any calls in such regions. Finally, we have developed CellMeSH - a new web server and API package for automatic cell-type identification in single-cell RNA-seq (scRNA-seq) analysis. CellMeSH predicts cell types, based on a set of marker genes as query input. CellMeSH builds its database in a scalable and easy-to-update way using prior literature, and adopts a novel probabilistic method to better query the database. Through a variety of experiments on human and mouse scRNA-seq datasets, CellMeSH has demonstrated richer gene and cell-type information in its database, robust query method, and an overall superior annotation performance.