148 research outputs found
GFF Utilities: GffRead and GffCompare
No full textclassify, merge, tracking and annotation of GFF files by comparing to a reference annotation GFFPlease cite as:
Pertea G and Pertea M. "GFF Utilities: GffRead and GffCompare". F1000Research 2020, 9:304 DOI: 10.12688/f1000research.23297.
GffCompare
No full textclassify, merge, tracking and annotation of GFF files by comparing to a reference annotation GFFPlease cite as: Pertea G and Pertea M. "GFF Utilities: GffRead and GffCompare". F1000Research 2020, 9:304 DOI: 10.12688/f1000research.23297.
eVIP2: Expression-based variant impact phenotyping to predict the function of gene variants
No full textWhile advancements in genome sequencing have identified millions of somatic mutations in cancer, their functional impact is poorly understood. We previously developed the expression-based variant impact phenotyping (eVIP) method to use gene expression data to characterize the function of gene variants. The eVIP method uses a decision tree-based algorithm to predict the functional impact of somatic variants by comparing gene expression signatures induced by introduction of wild-type (WT) versus mutant cDNAs in cell lines. The method distinguishes between variants that are gain-of-function, loss-of-function, change-of-function, or neutral. We present eVIP2, software that allows for pathway analysis (eVIP Pathways) and usage with RNA-seq data. To demonstrate the eVIP2 software and approach, we characterized two recurrent frameshift variants in RNF43, a negative regulator of Wnt signaling, frequently mutated in colorectal, gastric, and endometrial cancer. RNF43 WT, RNF43 R117fs, RNF43 G659fs, or GFP control cDNA were overexpressed in HEK293T cells. Analysis with eVIP2 predicted that the frameshift at position 117 was a loss-of-function mutation, as expected. The second frameshift at position 659 has been previously described as a passenger mutation that maintains the RNF43 WT function as a negative regulator of Wnt. Surprisingly, eVIP2 predicted G659fs to be a change-of-function mutation. Additional eVIP Pathways analysis of RNF43 G659fs predicted 10 pathways to be significantly altered, including TNF-α via NFκB signaling, KRAS signaling, and hypoxia, highlighting the benefit of a more comprehensive approach when determining the impact of gene variant function. To validate these predictions, we performed reporter assays and found that each pathway activated by expression of RNF43 G659fs, but not expression of RNF43 WT, was identified as impacted by eVIP2, supporting that RNF43 G659fs is a change-of-function mutation and its effect on the identified pathways. Pathway activation was further validated by Western blot analysis. Lastly, we show primary colon adenocarcinoma patient samples with R117fs and G659fs variants have transcriptional profiles similar to BRAF missense mutations with activated RAS/MAPK signaling, consistent with KRAS signaling pathways being GOF in both variants. The eVIP2 method is an important step towards overcoming the current challenge of variant interpretation in the implementation of precision medicine. eVIP2 is available at https://github.com/BrooksLabUCSC/eVIP2
ForestQC: Quality control on genetic variants from next-generation sequencing data using random forest
No full textNext-generation sequencing technology (NGS) enables the discovery of nearly all genetic variants present in a genome. A subset of these variants, however, may have poor sequencing quality due to limitations in NGS or variant callers. In genetic studies that analyze a large number of sequenced individuals, it is critical to detect and remove those variants with poor quality as they may cause spurious findings. In this paper, we present ForestQC, a statistical tool for performing quality control on variants identified from NGS data by combining a traditional filtering approach and a machine learning approach. Our software uses the information on sequencing quality, such as sequencing depth, genotyping quality, and GC contents, to predict whether a particular variant is likely to be false-positive. To evaluate ForestQC, we applied it to two whole-genome sequencing datasets where one dataset consists of related individuals from families while the other consists of unrelated individuals. Results indicate that ForestQC outperforms widely used methods for performing quality control on variants such as VQSR of GATK by considerably improving the quality of variants to be included in the analysis. ForestQC is also very efficient, and hence can be applied to large sequencing datasets. We conclude that combining a machine learning algorithm trained with sequencing quality information and the filtering approach is a practical approach to perform quality control on genetic variants from sequencing data.</div
Development of methods for the analysis of deep sequencing data; applications to the discovery of functions of RNA-binding proteins
No full textWith the recent advances in nucleotide sequencing technologies, it became easy to generate tens of millions of reads with genome- or transcriptome-wide distribution with reduced cost and high accuracy. One of the applications of deep sequencing is the determination of the repertoire of targets of RNA-binding proteins. The method, called CLIP (for UV crosslinking and immune-precipitation) is now widely used to characterize a variety of proteins with regulatory as well as enzymatic functions. Here we focus on the statistical analysis of data obtained through a variant of CLIP, called PAR-CLIP (Photoactivatable-Ribonucleoside-Enhanced CLIP), which was applied to three different RNA binding proteins whose function was previously not well characterized: PAPD5 (PAP associated domain containing 5), DIS3L2 (DIS3 mitotic control homolog (S. cerevisiae)-like 2), and EWSR1 (Ewing sarcoma breakpoint region 1). Our computational analysis was instrumental for the definition of the main in vivo substrates of these proteins, which were confirmed by additional experiments. In the analysis, we also used extensively publicly available high-throughput data sets that enabled us make inferences about the function of the proteins. The main results of biological significance were as follows. We determined ribosomal RNAs are the main targets of PAPD5 and that the main substrates of the DIS3L2 nuclease are tRNAs and found that the tRNA-derived fragments processed by DIS3L2 could be loaded in the RNA silencing complex and be involved in gene silencing. Finally, we determined that EWSR1preferentially binds to RNAs that originate from instability-prone regions like sub-telomeres, known to be hotspots of genomic rearrangements, as well as other genes located in internal regions of chromosomes, that have been implicated in genomic translocations. These include EWSR1’s own pre-mRNA. All together this dissertation illustrates the point that when coupled with proper statistical analysis, CLIP is able to reveal targets of RNA-binding proteins that were difficult to study with other methods and that and integration of public domain datasets is very powerful in deciphering complex RNA-protein and regulatory RNA networks implicated in post-transcriptional gene regulation
COMPUTATIONAL STUDY OF TRANSCRIPTIONAL LANDSCAPES FROM RNA-SEQ DATA
No full textSince the inception of genetic science in the days of Gregor Johann Mendel, the major focus of genetics has been to identify functional units, genes, passed through generations and to determine how variation affects the development of an organism. While genome projects allowed us to collect complete and accurate DNA sequences of individual organisms, novel functional elements are being routinely discovered. When RNA sequencing technologies appeared in the late 2000s, they opened a new view over all transcriptional activity of cells at varying levels of resolution. However, the imperfections of both technical and biological processes, the growing amounts of data as well as the overall complexity of eukaryotic genomes underscore the need for novel analytical approaches to discover new and improve understanding of known genes and isoforms.
This work begins with an overview of the status of human gene annotation and presents a comprehensive discussion of existing and new methods for creating complete and accurate gene catalogs, via comparative genomics and RNA sequencing. Through careful analysis of large RNA-seq datasets, we annotate effects of transcriptional artifacts and inaccuracies in gene expression on downstream analysis. Moreover, we identify defining properties of validated transcription that distinguish it from the effervescent noise. To address the challenges presented by noisy transcription and improve gene expression analysis, we introduce TieBrush, a comprehensive suite of tools for efficient processing of multi-sample sequencing datasets. This suite allows for the creation of condensed representations of data, facilitating the identification of shared transcriptional motifs and enhancing downstream analysis. Furthermore, to enhance our understanding of alternative splicing and distinguish functional isoforms from noise at protein-coding loci, we develop ORFanage. This highly efficient and accurate system assigns open reading frames (ORFs) to gene transcripts, thereby improving gene annotations. Finally, we employ all presented techniques to design a complete sample-to-annotation protocol for annotating genes and transcripts. We apply this protocol to create CHESS 3 - an improved human genome annotation, identifying multiple novel tissue specific isoforms while increasing consistency and reliability of known transcript models.
In addition to advancements in gene annotation, this thesis briefly explores the critical role of large representative datasets of viral genomes in acquiring novel insights into diseases. More specifically, we discuss how pangenomic analysis of HIV-1 facilitated novel insights into viral persistence and how the challenges of mass sequencing of SARS-CoV-2 genomes required a novel approach for identification of first emerging recombinant lineages
NOVEL METHODS IMPROVE GENOME ANNOTATION
No full textIn the era of high-throughput sequencing, comprehensive genome annotation has become critical for understanding the functional complexities of life. Here we explore the development and application of computational methods for genome annotation, focusing on human transcriptome analysis, prokaryotic gene prediction, and circular RNA annotation. The primary contributions of this work span three distinct areas, each addressing unique challenges in the field. First, we develop a structure-guided isoform identification approach that utilizes three-dimensional protein structure predictions to identify functional human gene isoforms. Our method evaluates over 230,000 isoforms of human protein-coding genes assembled from thousands of RNA sequencing experiments across various human tissues. We identify hundreds of isoforms with more confidently predicted structure and potentially superior function compared to canonical isoforms, thus demonstrating the potential of protein structure prediction as a powerful tool for genome annotation and transcriptome analysis. Second, we present a universal protein model for prokaryotic gene prediction, Balrog, which employs a temporal convolutional network to analyze amino acid sequences from a diverse set of microbial genomes. Balrog eliminates the need for genome-specific training and matches or outperforms existing state-of-the-art gene finding tools. Lastly, we introduce alignment-free methods for annotating circular RNA in humans, leveraging a simple k-mer-based data structure
COMPUTATIONAL STUDY OF TRANSCRIPTIONAL LANDSCAPES FROM RNA-SEQ DATA
No full textSince the inception of genetic science in the days of Gregor Johann Mendel, the major focus of genetics has been to identify functional units, genes, passed through generations and to determine how variation affects the development of an organism. While genome projects allowed us to collect complete and accurate DNA sequences of individual organisms, novel functional elements are being routinely discovered. When RNA sequencing technologies appeared in the late 2000s, they opened a new view over all transcriptional activity of cells at varying levels of resolution. However, the imperfections of both technical and biological processes, the growing amounts of data as well as the overall complexity of eukaryotic genomes underscore the need for novel analytical approaches to discover new and improve understanding of known genes and isoforms.
This work begins with an overview of the status of human gene annotation and presents a comprehensive discussion of existing and new methods for creating complete and accurate gene catalogs, via comparative genomics and RNA sequencing. Through careful analysis of large RNA-seq datasets, we annotate effects of transcriptional artifacts and inaccuracies in gene expression on downstream analysis. Moreover, we identify defining properties of validated transcription that distinguish it from the effervescent noise. To address the challenges presented by noisy transcription and improve gene expression analysis, we introduce TieBrush, a comprehensive suite of tools for efficient processing of multi-sample sequencing datasets. This suite allows for the creation of condensed representations of data, facilitating the identification of shared transcriptional motifs and enhancing downstream analysis. Furthermore, to enhance our understanding of alternative splicing and distinguish functional isoforms from noise at protein-coding loci, we develop ORFanage. This highly efficient and accurate system assigns open reading frames (ORFs) to gene transcripts, thereby improving gene annotations. Finally, we employ all presented techniques to design a complete sample-to-annotation protocol for annotating genes and transcripts. We apply this protocol to create CHESS 3 - an improved human genome annotation, identifying multiple novel tissue specific isoforms while increasing consistency and reliability of known transcript models.
In addition to advancements in gene annotation, this thesis briefly explores the critical role of large representative datasets of viral genomes in acquiring novel insights into diseases. More specifically, we discuss how pangenomic analysis of HIV-1 facilitated novel insights into viral persistence and how the challenges of mass sequencing of SARS-CoV-2 genomes required a novel approach for identification of first emerging recombinant lineages
The Human Transcriptome: An Unfinished Story
No full textDespite recent technological advances, the study of the human transcriptome is still in its early stages. Here we provide an overview of the complex human transcriptomic landscape, present the bioinformatics challenges posed by the vast quantities of transcriptomic data, and discuss some of the studies that have tried to determine how much of the human genome is transcribed. Recent evidence has suggested that more than 90% of the human genome is transcribed into RNA. However, this view has been strongly contested by groups of scientists who argued that many of the observed transcripts are simply the result of transcriptional noise. In this review, we conclude that the full extent of transcription remains an open question that will not be fully addressed until we decipher the complete range and biological diversity of the transcribed genomic sequences
- …
