Oarfish: Enhanced probabilistic modeling leads to improved accuracy in long read transcriptome quantification

Motivation

Long read sequencing technology is becoming an increasingly indispensable tool in genomic and transcriptomic analysis. In transcriptomics in particular, long reads offer the possibility of sequencing full-length isoforms, which can vastly simplify the identification of novel transcripts and transcript quantification. However, despite this promise, the focus of much long read method development to date has been on transcript identification, with comparatively little attention paid to quantification. Yet, due to differences in the underlying protocols and technologies, lower throughput (i.e. fewer reads sequenced per sample compared to short read technologies), as well as technical artifacts, long read quantification remains a challenge, motivating the continued development and assessment of quantification methods tailored to this increasingly prevalent type of data.

Results

We introduce a new method and software tool for long read transcript quantification called oarfish. Our model incorporates a novel and innovative coverage score, which affects the conditional probability of fragment assignment in the underlying probabilistic model. We demonstrate that by accounting for this coverage information, oarfish is able to produce more accurate quantification estimates than existing long read quantification methods, particularly when one considers the primary isoforms present in a particular cell line or tissue type.

Availability and Implementation

Oarfish is implemented in the Rust programming language, and is made available as free and open-source software under the BSD 3-clause license. The source code is available at https://www.github.com/COMBINE-lab/oarfish.

Context-aware transcript quantification from long-read RNA-seq data with Bambu

Most approaches to transcript quantification rely on fixed reference annotations; however, the transcriptome is dynamic and depending on the context, such static annotations contain inactive isoforms for some genes, whereas they are incomplete for others. Here we present Bambu, a method that performs machine-learning-based transcript discovery to enable quantification specific to the context of interest using long-read RNA-sequencing. To identify novel transcripts, Bambu estimates the novel discovery rate, which replaces arbitrary per-sample thresholds with a single, interpretable, precision-calibrated parameter. Bambu retains the full-length and unique read counts, enabling accurate quantification in presence of inactive isoforms. Compared to existing methods for transcript discovery, Bambu achieves greater precision without sacrificing sensitivity. We show that context-aware annotations improve quantification for both novel and known transcripts. We apply Bambu to quantify isoforms from repetitive HERVH-LTR7 retrotransposons in human embryonic stem cells, demonstrating the ability for context-specific transcript expression analysis.

Loss of HCN2 in Dorsal Hippocampus of Young Adult Mice Induces Specific Apoptosis of the CA1 Pyramidal Neuron Layer

Neurons inevitably rely on a proper repertoire and distribution of membrane-bound ion-conducting channels. Among these proteins, the family of hyperpolarization-activated and cyclic nucleotide-gated (HCN) channels possesses unique properties giving rise to the corresponding I_h-current that contributes to various aspects of neural signaling. In mammals, four genes (hcn1-4) encode subunits of HCN channels. These subunits can assemble as hetero- or homotetrameric ion-conducting channels. In order to elaborate on the specific role of the HCN2 subunit in shaping electrical properties of neurons, we applied an Adeno-associated virus (AAV)-mediated, RNAi-based knock-down strategy of hcn2 gene expression both in vitro and in vivo. Electrophysiological measurements showed that HCN2 subunit knock-down resulted in specific yet anticipated changes in I_h-current properties in primary hippocampal neurons and, in addition, corroborated that the HCN2 subunit participates in postsynaptic signal integration. To further address the role of the HCN2 subunit in vivo, we injected recombinant (r)AAVs into the dorsal hippocampus of young adult male mice. Behavioral and biochemical analyses were conducted to assess the contribution of HCN2-containing channels in shaping hippocampal network properties. Surprisingly, knock-down of hcn2 expression resulted in a severe degeneration of the CA1 pyramidal cell layer, which did not occur in mice injected with control rAAV constructs. This finding might pinpoint to a vital and yet unknown contribution of HCN2 channels in establishing or maintaining the proper function of CA1 pyramidal neurons of the dorsal hippocampus.

AAV-Mediated CRISPRi and RNAi Based Gene Silencing in Mouse Hippocampal Neurons

Uncovering the physiological role of individual proteins that are part of the intricate process of cellular signaling is often a complex and challenging task. A straightforward strategy of studying a protein's function is by manipulating the expression rate of its gene. In recent years, the Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR)/Cas9-based technology was established as a powerful gene-editing tool for generating sequence specific changes in proliferating cells. However, obtaining homogeneous populations of transgenic post-mitotic neurons by CRISPR/Cas9 turned out to be challenging. These constraints can be partially overcome by CRISPR interference (CRISPRi), which mediates the inhibition of gene expression by competing with the transcription machinery for promoter binding and, thus, transcription initiation. Notably, CRISPR/Cas is only one of several described approaches for the manipulation of gene expression. Here, we targeted neurons with recombinant Adeno-associated viruses to induce either CRISPRi or RNA interference (RNAi), a well-established method for impairing de novo protein biosynthesis by using cellular regulatory mechanisms that induce the degradation of pre-existing mRNA. We specifically targeted hyperpolarization-activated and cyclic nucleotide-gated (HCN) channels, which are widely expressed in neuronal tissues and play essential physiological roles in maintaining biophysical characteristics in neurons. Both of the strategies reduced the expression levels of three HCN isoforms (HCN1, 2, and 4) with high specificity. Furthermore, detailed analysis revealed that the knock-down of just a single HCN isoform (HCN4) in hippocampal neurons did not affect basic electrical parameters of transduced neurons, whereas substantial changes emerged in HCN-current specific properties.

AP-TSS: A New Method for the Analysis of RNA Expression from Particular and Challenging Transcription Start Sites

Alternative promoter usage involved in the regulation of transcription, splicing, and translation contributes to proteome diversity and is involved in a large number of diseases, in particular, cancer. Epigenetic mechanisms and cis regulatory elements are involved in alternative promoter activity. Multiple transcript isoforms can be produced from a gene, due to the initiation of transcription at different transcription start sites (TSS). These transcripts may not have regions that allow discrimination during RT-qPCR, making quantification technically challenging. This study presents a general method for the relative quantification of a transcript synthesized from a particular TSS that we called AP-TSS (analysis of particular TSS). AP-TSS is based on the specific elongation of the cDNA of interest, followed by its quantification by qPCR. As proof of principle, AP-TSS was applied to two non-coding RNA: telomeric repeat-containing RNAs (TERRA) from a particular subtelomeric TSS, and Alu transcripts. The treatment of cells with a DNA methylation inhibitor was associated with a global increase of the total TERRA level, but the TERRA expression from the TSS of interest did not change in HT1080 cells, and only modestly increased in HeLa cells. This result suggests that TERRA upregulation induced by global demethylation of the genome is mainly due to activation from sites other than this particular TSS. For Alu RNA, the signal obtained by AP-TSS is specific for the RNA Polymerase III-dependent Alu transcript. In summary, our method provides a tool to study regulation of gene expression from a given transcription start site, in different conditions that could be applied to many genes. In particular, AP-TSS can be used to investigate the epigenetic regulation of alternative TSS usage that is of importance for the development of epigenetic-targeted therapies.

Piecing the puzzle together: a revisit to transcript reconstruction problem in RNA-seq

The advancement of RNA sequencing (RNA-seq) has provided an unprecedented opportunity to assess both the diversity and quantity of transcript isoforms in an mRNA transcriptome. In this paper, we revisit the computational problem of transcript reconstruction and quantification. Unlike existing methods which focus on how to explain the exons and splice variants detected by the reads with a set of isoforms, we aim at reconstructing transcripts by piecing the reads into individual effective transcript copies. Simultaneously, the quantity of each isoform is explicitly measured by the number of assembled effective copies, instead of estimated solely based on the collective read count. We have developed a novel method named Astroid that solves the problem of effective copy reconstruction on the basis of a flow network. The RNA-seq reads are represented as vertices in the flow network and are connected by weighted edges that evaluate the likelihood of two reads originating from the same effective copy. A maximum likelihood set of transcript copies is then reconstructed by solving a minimum-cost flow problem on the flow network. Simulation studies on the human transcriptome have demonstrated the superior sensitivity and specificity of Astroid in transcript reconstruction as well as improved accuracy in transcript quantification over several existing approaches. The application of Astroid on two real RNA-seq datasets has further demonstrated its accuracy through high correlation between the estimated isoform abundance and the qRT-PCR validations.

Digital Droplet PCR: CNV Analysis and Other Applications

Digital droplet PCR (ddPCR) is an assay that combines state-of-the-art microfluidics technology with TaqMan-based PCR to achieve precise target DNA quantification at high levels of sensitivity and specificity. Because quantification is achieved without the need for standard assays in an easy to interpret, unambiguous digital readout, ddPCR is far simpler, faster, and less error prone than real-time qPCR. The basic protocol can be modified with minor adjustments to suit a wide range of applications, such as CNV analysis, rare variant detection, SNP genotyping, and transcript quantification. This unit describes the ddPCR workflow in detail for the Bio-Rad QX100 system, but the theory and data interpretation are generalizable to any ddPCR system.

Quantitative imaging of single mRNA splice variants in living cells

Alternative messenger RNA (mRNA) splicing is a fundamental process of gene regulation, and errors in RNA splicing are known to be associated with a variety of different diseases. However, there is currently a lack of quantitative technologies for monitoring mRNA splice variants in cells. Here, we show that a combination of plasmonic dimer probes and hyperspectral imaging can be used to detect and quantify mRNA splice variants in living cells. The probes are made from gold nanoparticles functionalized with oligonucleotides and can hybridize to specific mRNA sequences, forming nanoparticle dimers that exhibit distinct spectral shifts due to plasmonic coupling. With this approach, we show that the spatial and temporal distribution of three selected splice variants of the breast cancer susceptibility gene, BRCA1, can be monitored at single-copy resolution by measuring the hybridization dynamics of the nanoplasmonic dimers. Our study provides insights into RNA and its transport in living cells, which could improve our understanding of cellular protein complexes, pharmacogenomics, genetic diagnosis and gene therapies.