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

Contrasting and combining transcriptome complexity captured by short and long RNA sequencing reads

Mapping transcriptomic variations using either short- or long-read RNA sequencing is a staple of genomic research. Long reads are able to capture entire isoforms and overcome repetitive regions, whereas short reads still provide improved coverage and error rates. Yet, open questions remain, such as how to quantitatively compare the technologies, can we combine them, and what is the benefit of such a combined view? We tackle these questions by first creating a pipeline to assess matched long- and short-read data using a variety of transcriptome statistics. We find that across data sets, algorithms, and technologies, matched short-read data detects ∼30% more splice junctions, such that ∼10%-30% of the splice junctions included at ≥20% by short reads are missed by long reads. In contrast, long reads detect many more intron-retention events and can detect full isoforms, pointing to the benefit of combining the technologies. We introduce MAJIQ-L, an extension of the MAJIQ software, to enable a unified view of transcriptome variations from both technologies and demonstrate its benefits. Our software can be used to assess any future long-read technology or algorithm and can be combined with short-read data for improved transcriptome analysis.

Characterization and implementation of the MarathonRT template-switching reaction to expand the capabilities of RNA-seq

End-to-end RNA-sequencing methods that capture 5'-sequence content without cumbersome library manipulations are of great interest, particularly for analysis of long RNAs. While template-switching methods have been developed for RNA sequencing by distributive short-read RTs, such as the MMLV RTs used in SMART-Seq methods, they have not been adapted to leverage the power of ultraprocessive RTs, such as those derived from group II introns. To facilitate this transition, we dissected the individual processes that guide the enzymatic specificity and efficiency of the multistep template-switching reaction carried out by RTs, in this case, by MarathonRT. Remarkably, this is the first study of its kind, for any RT. First, we characterized the nucleotide specificity of nontemplated addition (NTA) reaction that occurs when the RT extends past the RNA 5'-terminus. We then evaluated the binding specificity of specialized template-switching oligonucleotides, optimizing their sequences and chemical properties to guide efficient template-switching reaction. Having dissected and optimized these individual steps, we then unified them into a procedure for performing RNA sequencing with MarathonRT enzymes, using a well-characterized RNA reference set. The resulting reads span a six-log range in transcript concentration and accurately represent the input RNA identities in both length and composition. We also performed RNA-seq from total human RNA and poly(A)-enriched RNA, with short- and long-read sequencing demonstrating that MarathonRT enhances the discovery of unseen RNA molecules by conventional RT. Altogether, we have generated a new pipeline for rapid, accurate sequencing of complex RNA libraries containing mixtures of long RNA transcripts.

Machine learning and related approaches in transcriptomics

Data acquisition for transcriptomic studies used to be the bottleneck in the transcriptomic analytical pipeline. However, recent developments in transcriptome profiling technologies have increased researchers' ability to obtain data, resulting in a shift in focus to data analysis. Incorporating machine learning to traditional analytical methods allows the possibility of handling larger volumes of complex data more efficiently. Many bioinformaticians, especially those unfamiliar with ML in the study of human transcriptomics and complex biological systems, face a significant barrier stemming from their limited awareness of the current landscape of ML utilisation in this field. To address this gap, this review endeavours to introduce those individuals to the general types of ML, followed by a comprehensive range of more specific techniques, demonstrated through examples of their incorporation into analytical pipelines for human transcriptome investigations. Important computational aspects such as data pre-processing, task formulation, results (performance of ML models), and validation methods are encompassed. In hope of better practical relevance, there is a strong focus on studies published within the last five years, almost exclusively examining human transcriptomes, with outcomes compared with standard non-ML tools.

CyCoNP lncRNA establishes cis and trans RNA-RNA interactions to supervise neuron physiology

The combination of morphogenetic and transcription factors together with the synergic aid of noncoding RNAs and their cognate RNA binding proteins contribute to shape motor neurons (MN) identity. Here, we extend the noncoding perspective of human MN, by detailing the molecular and biological activity of CyCoNP (as Cytoplasmic Coordinator of Neural Progenitors) a highly expressed and MN-enriched human lncRNA. Through in silico prediction, in vivo RNA purification and loss of function experiments followed by RNA-sequencing, we found that CyCoNP sustains a specific neuron differentiation program, required for the physiology of both neuroblastoma cells and hiPSC-derived MN, which mainly involves miR-4492 and NCAM1 mRNA. We propose a novel lncRNA-mediated 'dual mode' of action, in which CyCoNP acts in trans as a classical RNA sponge by sequestering miR-4492 from its pro-neuronal targets, including NCAM1 mRNA, and at the same time it plays an additional role in cis by interacting with NCAM1 mRNA and regulating the availability and localization of the miR-4492 in its proximity. These data highlight novel insights into the noncoding RNA-mediated control of human neuron physiology and point out the importance of lncRNA-mediated interactions for the spatial distribution of regulatory molecules.

Long-read sequencing transcriptome quantification with lr-kallisto

RNA abundance quantification has become routine and affordable thanks to high-throughput "short-read" technologies that provide accurate molecule counts at the gene level. Similarly accurate and affordable quantification of definitive full-length, transcript isoforms has remained a stubborn challenge, despite its obvious biological significance across a wide range of problems. "Long-read" sequencing platforms now produce data-types that can, in principle, drive routine definitive isoform quantification. However some particulars of contemporary long-read datatypes, together with isoform complexity and genetic variation, present bioinformatic challenges. We show here, using ONT data, that fast and accurate quantification of long-read data is possible and that it is improved by exome capture. To perform quantifications we developed lr-kallisto, which adapts the kallisto bulk and single-cell RNA-seq quantification methods for long-read technologies.

Discovery of NRG1-VII: the myeloid-derived class of NRG1

The growth factor Neuregulin-1 (NRG1) has pleiotropic roles in proliferation and differentiation of the stem cell niche in different tissues. It has been implicated in gut, brain and muscle development and repair. Six isoform classes of NRG1 and over 28 protein isoforms have been previously described. Here we report a new class of NRG1, designated NRG1-VII to denote that these NRG1 isoforms arise from a myeloid-specific transcriptional start site (TSS) previously uncharacterized. Long-read sequencing was used to identify eight high-confidence NRG1-VII transcripts. These transcripts presented major structural differences from one another, through the use of cassette exons and alternative stop codons. Expression of NRG1-VII was confirmed in primary human monocytes and tissue resident macrophages and induced pluripotent stem cell-derived macrophages (iPSC-derived macrophages). Isoform switching via cassette exon usage and alternate polyadenylation was apparent during monocyte maturation and macrophage differentiation. NRG1-VII is the major class expressed by the myeloid lineage, including tissue-resident macrophages. Analysis of public gene expression data indicates that monocytes and macrophages are a primary source of NRG1. The size and structure of class VII isoforms suggests that they may be more diffusible through tissues than other NRG1 classes. However, the specific roles of class VII variants in tissue homeostasis and repair have not yet been determined.

Accurate long-read transcript discovery and quantification at single-cell, pseudo-bulk and bulk resolution with Isosceles

Accurate detection and quantification of mRNA isoforms from nanopore long-read sequencing remains challenged by technical noise, particularly in single cells. To address this, we introduce Isosceles, a computational toolkit that outperforms other methods in isoform detection sensitivity and quantification accuracy across single-cell, pseudo-bulk and bulk resolution levels, as demonstrated using synthetic and biologically-derived datasets. Here we show Isosceles improves the fidelity of single-cell transcriptome quantification at the isoform-level, and enables flexible downstream analysis. As a case study, we apply Isosceles, uncovering coordinated splicing within and between neuronal differentiation lineages. Isosceles is suitable to be applied in diverse biological systems, facilitating studies of cellular heterogeneity across biomedical research applications.

Enhancing novel isoform discovery: leveraging nanopore long-read sequencing and machine learning approaches

Long-read sequencing technologies can capture entire RNA transcripts in a single sequencing read, reducing the ambiguity in constructing and quantifying transcript models in comparison to more common and earlier methods, such as short-read sequencing. Recent improvements in the accuracy of long-read sequencing technologies have expanded the scope for novel splice isoform detection and have also enabled a far more accurate reconstruction of complex splicing patterns and transcriptomes. Additionally, the incorporation and advancements of machine learning and deep learning algorithms in bioinformatic software have significantly improved the reliability of long-read sequencing transcriptomic studies. However, there is a lack of consensus on what bioinformatic tools and pipelines produce the most precise and consistent results. Thus, this review aims to discuss and compare the performance of available methods for novel isoform discovery with long-read sequencing technologies, with 25 tools being presented. Furthermore, this review intends to demonstrate the need for developing standard analytical pipelines, tools, and transcript model conventions for novel isoform discovery and transcriptomic studies.

Enhancing transcriptome expression quantification through accurate assignment of long RNA sequencing reads with TranSigner

Recently developed long-read RNA sequencing technologies promise to provide a more accurate and comprehensive view of transcriptomes compared to short-read sequencers, primarily due to their capability to achieve full-length sequencing of transcripts. However, realizing this potential requires computational tools tailored to process long reads, which exhibit a higher error rate than short reads. Existing methods for assembling and quantifying long-read data often disagree on expressed transcripts and their abundance levels, leading researchers to lack confidence in the transcriptomes produced using this data. One approach to address the uncertainties in transcriptome assembly and quantification is by assigning the long reads to transcripts, enabling a more detailed characterization of transcript support at the read level. Here, we introduce TranSigner, a versatile tool that assigns long reads to any input transcriptome. TranSigner consists of three consecutive modules performing: read alignment to the given transcripts, computation of read-to-transcript compatibility based on alignment scores and positions, and execution of an expectation-maximization algorithm to probabilistically assign reads to transcripts and estimate transcript abundances. Using simulated data and experimental datasets from three well-studied organisms - Homo sapiens, Arabidopsis thaliana, and Mus musculus - we demonstrate that TranSigner achieves accurate read assignments, obtaining higher accuracy in transcript abundance estimation compared to existing tools.

Nanopore guided annotation of transcriptome architectures

Nanopore direct RNA sequencing (DRS) enables the capture and full-length sequencing of native RNAs, without recoding or amplification bias. Resulting data sets may be interrogated to define the identity and location of chemically modified ribonucleotides, as well as the length of poly(A) tails, on individual RNA molecules. The success of these analyses is highly dependent on the provision of high-resolution transcriptome annotations in combination with workflows that minimize misalignments and other analysis artifacts. Existing software solutions for generating high-resolution transcriptome annotations are poorly suited to small gene-dense genomes of viruses due to the challenge of identifying distinct transcript isoforms where alternative splicing and overlapping RNAs are prevalent. To resolve this, we identified key characteristics of DRS data sets that inform resulting read alignments and developed the nanopore guided annotation of transcriptome architectures (NAGATA) software package (https://github.com/DepledgeLab/NAGATA). We demonstrate, using a combination of synthetic and original DRS data sets derived from adenoviruses, herpesviruses, coronaviruses, and human cells, that NAGATA outperforms existing transcriptome annotation software and yields a consistently high level of precision and recall when reconstructing both gene sparse and gene-dense transcriptomes. Finally, we apply NAGATA to generate the first high-resolution transcriptome annotation of the neglected pathogen human adenovirus type F41 (HAdV-41) for which we identify 77 distinct transcripts encoding at least 23 different proteins.

IMPORTANCE

The transcriptome of an organism denotes the full repertoire of encoded RNAs that may be expressed. This is critical to understanding the biology of an organism and for accurate transcriptomic and epitranscriptomic-based analyses. Annotating transcriptomes remains a complex task, particularly in small gene-dense organisms such as viruses which maximize their coding capacity through overlapping RNAs. To resolve this, we have developed a new software nanopore guided annotation of transcriptome architectures (NAGATA) which utilizes nanopore direct RNA sequencing (DRS) datasets to rapidly produce high-resolution transcriptome annotations for diverse viruses and other organisms.

Accurate isoform quantification by joint short- and long-read RNA-sequencing

Accurate quantification of transcript isoforms is crucial for understanding gene regulation, functional diversity, and cellular behavior. Existing RNA sequencing methods have significant limitations: short-read (SR) sequencing provides high depth but struggles with isoform deconvolution, whereas long-read (LR) sequencing offers isoform resolution at the cost of lower depth, higher noise, and technical biases. Addressing this gap, we introduce Multi-Platform Aggregation and Quantification of Transcripts (MPAQT), a generative model that combines the complementary strengths of different sequencing platforms to achieve state-of-the-art isoform-resolved transcript quantification, as demonstrated by extensive simulations and experimental benchmarks. By applying MPAQT to an in vitro model of human embryonic stem cell differentiation into cortical neurons, followed by machine learning-based modeling of transcript abundances, we show that untranslated regions (UTRs) are major determinants of isoform proportion and exon usage; this effect is mediated through isoform-specific sequence features embedded in UTRs, which likely interact with RNA-binding proteins that modulate mRNA stability. These findings highlight MPAQT's potential to enhance our understanding of transcriptomic complexity and underline the role of splicing-independent post-transcriptional mechanisms in shaping the isoform and exon usage landscape of the cell.

Detecting haplotype-specific transcript variation in long reads with FLAIR2

BACKGROUND

RNA-seq has brought forth significant discoveries regarding aberrations in RNA processing, implicating these RNA variants in a variety of diseases. Aberrant splicing and single nucleotide variants (SNVs) in RNA have been demonstrated to alter transcript stability, localization, and function. In particular, the upregulation of ADAR, an enzyme that mediates adenosine-to-inosine editing, has been previously linked to an increase in the invasiveness of lung adenocarcinoma cells and associated with splicing regulation. Despite the functional importance of studying splicing and SNVs, the use of short-read RNA-seq has limited the community's ability to interrogate both forms of RNA variation simultaneously.

RESULTS

We employ long-read sequencing technology to obtain full-length transcript sequences, elucidating cis-effects of variants on splicing changes at a single molecule level. We develop a computational workflow that augments FLAIR, a tool that calls isoform models expressed in long-read data, to integrate RNA variant calls with the associated isoforms that bear them. We generate nanopore data with high sequence accuracy from H1975 lung adenocarcinoma cells with and without knockdown of ADAR. We apply our workflow to identify key inosine isoform associations to help clarify the prominence of ADAR in tumorigenesis.

CONCLUSIONS

Ultimately, we find that a long-read approach provides valuable insight toward characterizing the relationship between RNA variants and splicing patterns.

Systematic assessment of long-read RNA-seq methods for transcript identification and quantification

The Long-read RNA-Seq Genome Annotation Assessment Project Consortium was formed to evaluate the effectiveness of long-read approaches for transcriptome analysis. Using different protocols and sequencing platforms, the consortium generated over 427 million long-read sequences from complementary DNA and direct RNA datasets, encompassing human, mouse and manatee species. Developers utilized these data to address challenges in transcript isoform detection, quantification and de novo transcript detection. The study revealed that libraries with longer, more accurate sequences produce more accurate transcripts than those with increased read depth, whereas greater read depth improved quantification accuracy. In well-annotated genomes, tools based on reference sequences demonstrated the best performance. Incorporating additional orthogonal data and replicate samples is advised when aiming to detect rare and novel transcripts or using reference-free approaches. This collaborative study offers a benchmark for current practices and provides direction for future method development in transcriptome analysis.

Surveying the landscape of RNA isoform diversity and expression across 9 GTEx tissues using long-read sequencing data

Even though alternative RNA splicing was discovered nearly 50 years ago (1977), we still understand very little about most isoforms arising from a single gene, including in which tissues they are expressed and if their functions differ. Human gene annotations suggest remarkable transcriptional complexity, with approximately 252,798 distinct RNA isoform annotations from 62,710 gene bodies (Ensembl v109; 2023), emphasizing the need to understand their biological effects. For example, 256 gene bodies have ≥50 annotated isoforms and 30 have ≥100, where one protein-coding gene (MAPK10) even has 192 distinct RNA isoform annotations. Whether such isoform diversity results from biological redundancy or spurious alternative splicing (i.e., noise), or whether individual isoforms have specialized functions (even if subtle) remains a mystery for most genes. Recent studies by Aguzzoli-Heberle et al., Leung et al., and Glinos et al. demonstrated long-read RNAseq enables improved RNA isoform quantification for essentially any tissue, cell type, or biological condition (e.g., disease, development, aging, etc.), making it possible to better assess individual isoform expression and function. While each study provided important discoveries related to RNA isoform diversity, deeper exploration is needed. We sought to quantify and characterize real isoform usage across tissues (compared to annotations). We used long-read RNAseq data from 58 GTEx samples across nine tissues (three brain, two heart, muscle, lung, liver, and cultured fibroblasts) generated by Glinos et al. and found considerable isoform diversity within and across tissues. Cerebellar hemisphere was the most transcriptionally complex tissue (22,522 distinct isoforms; 3,726 unique); liver was least diverse (12,435 distinct isoforms; 1,039 unique). We highlight gene clusters exhibiting high tissue-specific isoform diversity per tissue (e.g., TPM1 expresses 19 in heart's atrial appendage). We also validated 447 of the 700 new isoforms discovered by Aguzzoli-Heberle et al. and found that 88 were expressed in all nine tissues, while 58 were specific to a single tissue. This study represents a broad survey of the RNA isoform landscape, demonstrating isoform diversity across nine tissues and emphasizes the need to better understand how individual isoforms from a single gene body contribute to human health and disease.

Long-read RNA sequencing identifies region- and sex-specific C57BL/6J mouse brain mRNA isoform expression and usage

Alternative splicing (AS) contributes to the biological heterogeneity between species, sexes, tissues, and cell types. Many diseases are either caused by alterations in AS or by alterations to AS. Therefore, measuring AS accurately and efficiently is critical for assessing molecular phenotypes, including those associated with disease. Long-read sequencing enables more accurate quantification of differentially spliced isoform expression than short-read sequencing approaches, and third-generation platforms facilitate high-throughput experiments. To assess differences in AS across the cerebellum, cortex, hippocampus, and striatum by sex, we generated and analyzed Oxford Nanopore Technologies (ONT) long-read RNA sequencing (lrRNA-Seq) C57BL/6J mouse brain cDNA libraries. From > 85 million reads that passed quality control metrics, we calculated differential gene expression (DGE), differential transcript expression (DTE), and differential transcript usage (DTU) across brain regions and by sex. We found significant DGE, DTE, and DTU across brain regions and that the cerebellum had the most differences compared to the other three regions. Additionally, we found region-specific differential splicing between sexes, with the most sex differences in DTU in the cortex and no DTU in the hippocampus. We also report on two distinct patterns of sex DTU we observed, sex-divergent and sex-specific, that could potentially help explain sex differences in the prevalence and prognosis of various neurological and psychiatric disorders in future studies. Finally, we built a Shiny web application for researchers to explore the data further. Our study provides a resource for the community; it underscores the importance of AS in biological heterogeneity and the utility of long-read sequencing to better understand AS in the brain.

Isoform-specific RNA structure determination using Nano-DMS-MaP

RNA structure determination is essential to understand how RNA carries out its diverse biological functions. In cells, RNA isoforms are readily expressed with partial variations within their sequences due, for example, to alternative splicing, heterogeneity in the transcription start site, RNA processing or differential termination/polyadenylation. Nanopore dimethyl sulfate mutational profiling (Nano-DMS-MaP) is a method for in situ isoform-specific RNA structure determination. Unlike similar methods that rely on short sequencing reads, Nano-DMS-MaP employs nanopore sequencing to resolve the structures of long and highly similar RNA molecules to reveal their previously hidden structural differences. This Protocol describes the development and applications of Nano-DMS-MaP and outlines the main considerations for designing and implementing a successful experiment: from bench to data analysis. In cell probing experiments can be carried out by an experienced molecular biologist in 3-4 d. Data analysis requires good knowledge of command line tools and Python scripts and requires a further 3-5 d.

Mapping medically relevant RNA isoform diversity in the aged human frontal cortex with deep long-read RNA-seq

Determining whether the RNA isoforms from medically relevant genes have distinct functions could facilitate direct targeting of RNA isoforms for disease treatment. Here, as a step toward this goal for neurological diseases, we sequenced 12 postmortem, aged human frontal cortices (6 Alzheimer disease cases and 6 controls; 50% female) using one Oxford Nanopore PromethION flow cell per sample. We identified 1,917 medically relevant genes expressing multiple isoforms in the frontal cortex where 1,018 had multiple isoforms with different protein-coding sequences. Of these 1,018 genes, 57 are implicated in brain-related diseases including major depression, schizophrenia, Parkinson's disease and Alzheimer disease. Our study also uncovered 53 new RNA isoforms in medically relevant genes, including several where the new isoform was one of the most highly expressed for that gene. We also reported on five mitochondrially encoded, spliced RNA isoforms. We found 99 differentially expressed RNA isoforms between cases with Alzheimer disease and controls.

Epitranscriptome insights into Riccia fluitans L. (Marchantiophyta) aquatic transition using nanopore direct RNA sequencing

BACKGROUND

Riccia fluitans, an amphibious liverwort, exhibits a fascinating adaptation mechanism to transition between terrestrial and aquatic environments. Utilizing nanopore direct RNA sequencing, we try to capture the complex epitranscriptomic changes undergone in response to land-water transition.

RESULTS

A significant finding is the identification of 45 differentially expressed genes (DEGs), with a split of 33 downregulated in terrestrial forms and 12 upregulated in aquatic forms, indicating a robust transcriptional response to environmental changes. Analysis of N6-methyladenosine (m6A) modifications revealed 173 m6A sites in aquatic and only 27 sites in the terrestrial forms, indicating a significant increase in methylation in the former, which could facilitate rapid adaptation to changing environments. The aquatic form showed a global elongation bias in poly(A) tails, which is associated with increased mRNA stability and efficient translation, enhancing the plant's resilience to water stress. Significant differences in polyadenylation signals were observed between the two forms, with nine transcripts showing notable changes in tail length, suggesting an adaptive mechanism to modulate mRNA stability and translational efficiency in response to environmental conditions. This differential methylation and polyadenylation underline a sophisticated layer of post-transcriptional regulation, enabling Riccia fluitans to fine-tune gene expression in response to its living conditions.

CONCLUSIONS

These insights into transcriptome dynamics offer a deeper understanding of plant adaptation strategies at the molecular level, contributing to the broader knowledge of plant biology and evolution. These findings underscore the sophisticated post-transcriptional regulatory strategies Riccia fluitans employs to navigate the challenges of aquatic versus terrestrial living, highlighting the plant's dynamic adaptation to environmental stresses and its utility as a model for studying adaptation mechanisms in amphibious plants.

Comprehensive assessment of mRNA isoform detection methods for long-read sequencing data

The advancement of Long-Read Sequencing (LRS) techniques has significantly increased the length of sequencing to several kilobases, thereby facilitating the identification of alternative splicing events and isoform expressions. Recently, numerous computational tools for isoform detection using long-read sequencing data have been developed. Nevertheless, there remains a deficiency in comparative studies that systemically evaluate the performance of these tools, which are implemented with different algorithms, under various simulations that encompass potential influencing factors. In this study, we conducted a benchmark analysis of thirteen methods implemented in nine tools capable of identifying isoform structures from long-read RNA-seq data. We evaluated their performances using simulated data, which represented diverse sequencing platforms generated by an in-house simulator, RNA sequins (sequencing spike-ins) data, as well as experimental data. Our findings demonstrate IsoQuant as a highly effective tool for isoform detection with LRS, with Bambu and StringTie2 also exhibiting strong performance. These results offer valuable guidance for future research on alternative splicing analysis and the ongoing improvement of tools for isoform detection using LRS data.

Nanopore Guided Annotation of Transcriptome Architectures

High-resolution annotations of transcriptomes from all domains of life are essential for many sequencing-based RNA analyses, including Nanopore direct RNA sequencing (DRS), which would otherwise be hindered by misalignments and other analysis artefacts. DRS allows the capture and full-length sequencing of native RNAs, without recoding or amplification bias, and resulting data may be interrogated to define the identity and location of chemically modified ribonucleotides, as well as the length of poly(A) tails on individual RNA molecules. Existing software solutions for generating high-resolution transcriptome annotations are poorly suited to small gene dense organisms such as viruses due to the challenge of identifying distinct transcript isoforms where alternative splicing and overlapping RNAs are prevalent. To resolve this, we identified key characteristics of DRS datasets and developed a novel approach to transcriptome. We demonstrate, using a combination of synthetic and original datasets, that our novel approach yields a high level of precision and recall when reconstructing both gene sparse and gene dense transcriptomes from DRS datasets. We further apply this approach to generate a new high resolution transcriptome annotation of the neglected pathogen human adenovirus type F 41 for which we identify 77 distinct transcripts encoding at least 23 different proteins.

Single-cell and spatial transcriptomics: Bridging current technologies with long-read sequencing

Single-cell technologies have transformed biomedical research over the last decade, opening up new possibilities for understanding cellular heterogeneity, both at the genomic and transcriptomic level. In addition, more recent developments of spatial transcriptomics technologies have made it possible to profile cells in their tissue context. In parallel, there have been substantial advances in sequencing technologies, and the third generation of methods are able to produce reads that are tens of kilobases long, with error rates matching the second generation short reads. Long reads technologies make it possible to better map large genome rearrangements and quantify isoform specific abundances. This further improves our ability to characterize functionally relevant heterogeneity. Here, we show how researchers have begun to combine single-cell, spatial transcriptomics, and long-read technologies, and how this is resulting in powerful new approaches to profiling both the genome and the transcriptome. We discuss the achievements so far, and we highlight remaining challenges and opportunities.

Biosurfer for systematic tracking of regulatory mechanisms leading to protein isoform diversity

Long-read RNA sequencing has shed light on transcriptomic complexity, but questions remain about the functionality of downstream protein products. We introduce Biosurfer, a computational approach for comparing protein isoforms, while systematically tracking the transcriptional, splicing, and translational variations that underlie differences in the sequences of the protein products. Using Biosurfer, we analyzed the differences in 32,799 pairs of GENCODE annotated protein isoforms, finding a majority (70%) of variable N-termini are due to the alternative transcription start sites, while only 9% arise from 5' UTR alternative splicing. Biosurfer's detailed tracking of nucleotide-to-residue relationships helped reveal an uncommonly tracked source of single amino acid residue changes arising from the codon splits at junctions. For 17% of internal sequence changes, such split codon patterns lead to single residue differences, termed "ragged codons". Of variable C-termini, 72% involve splice- or intron retention-induced reading frameshifts. We found an unusual pattern of reading frame changes, in which the first frameshift is closely followed by a distinct second frameshift that restores the original frame, which we term a "snapback" frameshift. We analyzed long read RNA-seq-predicted proteome of a human cell line and found similar trends as compared to our GENCODE analysis, with the exception of a higher proportion of isoforms predicted to undergo nonsense-mediated decay. Biosurfer's comprehensive characterization of long-read RNA-seq datasets should accelerate insights of the functional role of protein isoforms, providing mechanistic explanation of the origins of the proteomic diversity driven by the alternative splicing. Biosurfer is available as a Python package at https://github.com/sheynkman-lab/biosurfer.

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.

Mis-spliced transcripts generate de novo proteins in TDP-43-related ALS/FTD

Functional loss of TDP-43, an RNA binding protein genetically and pathologically linked to amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), leads to the inclusion of cryptic exons in hundreds of transcripts during disease. Cryptic exons can promote the degradation of affected transcripts, deleteriously altering cellular function through loss-of-function mechanisms. Here, we show that mRNA transcripts harboring cryptic exons generated de novo proteins in TDP-43-depleted human iPSC-derived neurons in vitro, and de novo peptides were found in cerebrospinal fluid (CSF) samples from patients with ALS or FTD. Using coordinated transcriptomic and proteomic studies of TDP-43-depleted human iPSC-derived neurons, we identified 65 peptides that mapped to 12 cryptic exons. Cryptic exons identified in TDP-43-depleted human iPSC-derived neurons were predictive of cryptic exons expressed in postmortem brain tissue from patients with TDP-43 proteinopathy. These cryptic exons produced transcript variants that generated de novo proteins. We found that the inclusion of cryptic peptide sequences in proteins altered their interactions with other proteins, thereby likely altering their function. Last, we showed that 18 de novo peptides across 13 genes were present in CSF samples from patients with ALS/FTD spectrum disorders. The demonstration of cryptic exon translation suggests new mechanisms for ALS/FTD pathophysiology downstream of TDP-43 dysfunction and may provide a potential strategy to assay TDP-43 function in patient CSF.

Multiple Sgip1 splice variants inhibit cannabinoid receptor 1 internalization

Alternative splicing can often result in the expression of distinct protein isoforms from a single gene, with specific composition and properties. SH3-containing GRB2-like protein 3-interacting protein 1 (Sgip1) is a brain-enriched protein that regulates clathrin-mediated endocytosis and interferes with the internalization of cannabinoid receptor 1. Several research groups have studied the physiological importance of Sgip1, and four Sgip1 protein isoforms have been described to date, while the NCBI Gene database predicts the expression of 20 splice variants from the Sgip1 gene in mice. In this work, we cloned 15 Sgip1 splice variants from the mouse brain, including 11 novel splice variants. The cloned splice variants differed in exon composition within two Sgip1 regions: the membrane phospholipid-binding domain and the proline-rich region. All the Sgip1 splice isoforms had similar stability and comparable ability to inhibit the internalization of cannabinoid receptor 1. None of the isoforms influenced the internalization of the µ-opioid receptor. We confirm the expression of Sgip1 splice variants described in previous studies or predicted in silico. Our data provide a basis for further studies exploring the significance of Sgip1 splicing, and we suggest a new classification of Sgip1 splice variants to unify their nomenclature.

Long-read RNA sequencing identifies region- and sex-specific C57BL/6J mouse brain mRNA isoform expression and usage

Alternative splicing (AS) contributes to the biological heterogeneity between species, sexes, tissues, and cell types. Many diseases are either caused by alterations in AS or by alterations to AS. Therefore, measuring AS accurately and efficiently is critical for assessing molecular phenotypes, including those associated with disease. Long-read sequencing enables more accurate quantification of differentially spliced isoform expression than short-read sequencing approaches, and third-generation platforms facilitate high-throughput experiments. To assess differences in AS across the cerebellum, cortex, hippocampus, and striatum by sex, we generated and analyzed Oxford Nanopore Technologies (ONT) long-read RNA sequencing (lrRNA-Seq) C57BL/6J mouse brain cDNA libraries. From >85 million reads that passed quality control metrics, we calculated differential gene expression (DGE), differential transcript expression (DTE), and differential transcript usage (DTU) across brain regions and by sex. We found significant DGE, DTE, and DTU across brain regions and that the cerebellum had the most differences compared to the other three regions. Additionally, we found region-specific differential splicing between sexes, with the most sex differences in DTU in the cortex and no DTU in the hippocampus. We also report on two distinct patterns of sex DTU we observed, sex-divergent and sex-specific, that could potentially help explain sex differences in the prevalence and prognosis of various neurological and psychiatric disorders in future studies. Finally, we built a Shiny web application for researchers to explore the data further. Our study provides a resource for the community; it underscores the importance of AS in biological heterogeneity and the utility of long-read sequencing to better understand AS in the brain.

Alternative splicing induced by bacterial pore-forming toxins sharpens CIRBP-mediated cell response to Listeria infection

Cell autonomous responses to intracellular bacteria largely depend on reorganization of gene expression. To gain isoform-level resolution of these modes of regulation, we combined long- and short-read transcriptomic analyses of the response of intestinal epithelial cells to infection by the foodborne pathogen Listeria monocytogenes. Among the most striking isoform-based types of regulation, expression of the cellular stress response regulator CIRBP (cold-inducible RNA-binding protein) and of several SRSFs (serine/arginine-rich splicing factors) switched from canonical transcripts to nonsense-mediated decay-sensitive isoforms by inclusion of 'poison exons'. We showed that damage to host cell membranes caused by bacterial pore-forming toxins (listeriolysin O, perfringolysin, streptolysin or aerolysin) led to the dephosphorylation of SRSFs via the inhibition of the kinase activity of CLK1, thereby driving CIRBP alternative splicing. CIRBP isoform usage was found to have consequences on infection, since selective repression of canonical CIRBP reduced intracellular bacterial load while that of the poison exon-containing isoform exacerbated it. Consistently, CIRBP-bound mRNAs were shifted towards stress-relevant transcripts in infected cells, with increased mRNA levels or reduced translation efficiency for some targets. Our results thus generalize the alternative splicing of CIRBP and SRSFs as a common response to biotic or abiotic stresses by extending its relevance to the context of bacterial infection.

Using deep long-read RNAseq in Alzheimer's disease brain to assess medical relevance of RNA isoform diversity

Due to alternative splicing, human protein-coding genes average over eight RNA isoforms, resulting in nearly four distinct protein coding sequences per gene. Long-read RNAseq (IsoSeq) enables more accurate quantification of isoforms, shedding light on their specific roles. To assess the medical relevance of measuring RNA isoform expression, we sequenced 12 aged human frontal cortices (6 Alzheimer's disease cases and 6 controls; 50% female) using one Oxford Nanopore PromethION flow cell per sample. Our study uncovered 53 new high-confidence RNA isoforms in medically relevant genes, including several where the new isoform was one of the most highly expressed for that gene. Specific examples include WDR4 (61%; microcephaly), MYL3 (44%; hypertrophic cardiomyopathy), and MTHFS (25%; major depression, schizophrenia, bipolar disorder). Other notable genes with new high-confidence isoforms include CPLX2 (10%; schizophrenia, epilepsy) and MAOB (9%; targeted for Parkinson's disease treatment). We identified 1,917 medically relevant genes expressing multiple isoforms in human frontal cortex, where 1,018 had multiple isoforms with different protein coding sequences, demonstrating the need to better understand how individual isoforms from a single gene body are involved in human health and disease, if at all. Exactly 98 of the 1,917 genes are implicated in brain-related diseases, including Alzheimer's disease genes such as APP (Aβ precursor protein; five), MAPT (tau protein; four), and BIN1 (eight). As proof of concept, we also found 99 differentially expressed RNA isoforms between Alzheimer's cases and controls, despite the genes themselves not exhibiting differential expression. Our findings highlight the significant knowledge gaps in RNA isoform diversity and their medical relevance. Deep long-read RNA sequencing will be necessary going forward to fully comprehend the medical relevance of individual isoforms for a "single" gene.

SQANTI-SIM: a simulator of controlled transcript novelty for lrRNA-seq benchmark

Long-read RNA sequencing has emerged as a powerful tool for transcript discovery, even in well-annotated organisms. However, assessing the accuracy of different methods in identifying annotated and novel transcripts remains a challenge. Here, we present SQANTI-SIM, a versatile tool that wraps around popular long-read simulators to allow precise management of transcript novelty based on the structural categories defined by SQANTI3. By selectively excluding specific transcripts from the reference dataset, SQANTI-SIM effectively emulates scenarios involving unannotated transcripts. Furthermore, the tool provides customizable features and supports the simulation of additional types of data, representing the first multi-omics simulation tool for the lrRNA-seq field.

Long-read RNA-seq atlas of novel microglia isoforms elucidates disease-associated genetic regulation of splicing

Microglia, the innate immune cells of the central nervous system, have been genetically implicated in multiple neurodegenerative diseases. We previously mapped the genetic regulation of gene expression and mRNA splicing in human microglia, identifying several loci where common genetic variants in microglia-specific regulatory elements explain disease risk loci identified by GWAS. However, identifying genetic effects on splicing has been challenging due to the use of short sequencing reads to identify causal isoforms. Here we present the isoform-centric microglia genomic atlas (isoMiGA) which leverages the power of long-read RNA-seq to identify 35,879 novel microglia isoforms. We show that the novel microglia isoforms are involved in stimulation response and brain region specificity. We then quantified the expression of both known and novel isoforms in a multi-ethnic meta-analysis of 555 human microglia short-read RNA-seq samples from 391 donors, the largest to date, and found associations with genetic risk loci in Alzheimer's disease and Parkinson's disease. We nominate several loci that may act through complex changes in isoform and splice site usage.

Restrander: rapid orientation and artefact removal for long-read cDNA data

In transcriptomic analyses, it is helpful to keep track of the strand of the RNA molecules. However, the Oxford Nanopore long-read cDNA sequencing protocols generate reads that correspond to either the first or second-strand cDNA, therefore the strandedness of the initial transcript has to be inferred bioinformatically. Reverse transcription and PCR can also introduce artefacts which should be flagged in data pre-processing. Here we introduce Restrander, a lightning-fast and highly accurate tool for restranding and removing artefacts in long-read cDNA sequencing data. Thanks to its C++ implementation, Restrander was faster than Oxford Nanopore Technologies' existing tool Pychopper, and correctly restranded more reads due to its strategy of searching for polyA/T tails in addition to primer sequences from the reverse transcription and template-switch steps. We found that restranding improved the process of visualising and exploring data, and increased the number of novel isoforms discovered by bambu, particularly in regions where sense and anti-sense transcripts co-occur. The artefact detection implemented in Restrander quantifies reads lacking the correct 5' and 3' ends, a useful feature in quality control for library preparation. Restrander is pre-configured for all major cDNA protocols, and can be customised with user-defined primers. Restrander is available at https://github.com/mritchielab/restrander.

Adaptive sampling for nanopore direct RNA-sequencing

Nanopore long-read sequencing enables real-time monitoring and controlling of individual nanopores. This allows us to enrich or deplete specific sequences in DNA sequencing in a process called "adaptive sampling." So far, adaptive sampling (AS) was not applicable to the direct sequencing of RNA. Here, we show that AS is feasible and useful for direct RNA sequencing (DRS), which has its specific technical and biological challenges. Using a well-controlled in vitro transcript-based model system, we identify essential characteristics and parameter settings for AS in DRS, as the superior performance of depletion over enrichment. Here, the efficiency of depletion is close to the theoretical maximum. Additionally, we demonstrate that AS efficiently depletes specific transcripts in transcriptome-wide sequencing applications. Specifically, we applied our AS approach to poly(A)-enriched RNA samples from human-induced pluripotent stem cell-derived cardiomyocytes and mouse whole heart tissue and show efficient 2.5- to 2.8-fold depletion of highly abundant mitochondrial-encoded transcripts. Finally, we characterize depletion and enrichment performance for complex transcriptome subsets, that is, at the level of the entire Chromosome 11, proving the general applicability of direct RNA AS. Our analyses provide evidence that AS is especially useful to enable the detection of lowly expressed transcripts and reduce the sequencing of highly abundant disturbing transcripts.

Identification and quantification of small exon-containing isoforms in long-read RNA sequencing data

Small exons are pervasive in transcriptomes across organisms, and their quantification in RNA isoforms is crucial for understanding gene functions. Although long-read RNA-seq based on Oxford Nanopore Technologies (ONT) offers the advantage of covering transcripts in full length, its lower base accuracy poses challenges for identifying individual exons, particularly microexons (≤ 30 nucleotides). Here, we systematically assess small exons quantification in synthetic and human ONT RNA-seq datasets. We demonstrate that reads containing small exons are often not properly aligned, affecting the quantification of relevant transcripts. Thus, we develop a local-realignment method for misaligned exons (MisER), which remaps reads with misaligned exons to the transcript references. Using synthetic and simulated datasets, we demonstrate the high sensitivity and specificity of MisER for the quantification of transcripts containing small exons. Moreover, MisER enabled us to identify small exons with a higher percent spliced-in index (PSI) in neural, particularly neural-regulated microexons, when comparing 14 neural to 16 non-neural tissues in humans. Our work introduces an improved quantification method for long-read RNA-seq and especially facilitates studies using ONT long-reads to elucidate the regulation of genes involving small exons.

Inflammation primes the kidney for recovery by activating AZIN1 A-to-I editing

The progression of kidney disease varies among individuals, but a general methodology to quantify disease timelines is lacking. Particularly challenging is the task of determining the potential for recovery from acute kidney injury following various insults. Here, we report that quantitation of post-transcriptional adenosine-to-inosine (A-to-I) RNA editing offers a distinct genome-wide signature, enabling the delineation of disease trajectories in the kidney. A well-defined murine model of endotoxemia permitted the identification of the origin and extent of A-to-I editing, along with temporally discrete signatures of double-stranded RNA stress and Adenosine Deaminase isoform switching. We found that A-to-I editing of Antizyme Inhibitor 1 (AZIN1), a positive regulator of polyamine biosynthesis, serves as a particularly useful temporal landmark during endotoxemia. Our data indicate that AZIN1 A-to-I editing, triggered by preceding inflammation, primes the kidney and activates endogenous recovery mechanisms. By comparing genetically modified human cell lines and mice locked in either A-to-I edited or uneditable states, we uncovered that AZIN1 A-to-I editing not only enhances polyamine biosynthesis but also engages glycolysis and nicotinamide biosynthesis to drive the recovery phenotype. Our findings implicate that quantifying AZIN1 A-to-I editing could potentially identify individuals who have transitioned to an endogenous recovery phase. This phase would reflect their past inflammation and indicate their potential for future recovery.

Benchmarking long-read RNA-sequencing analysis tools using in silico mixtures

The lack of benchmark data sets with inbuilt ground-truth makes it challenging to compare the performance of existing long-read isoform detection and differential expression analysis workflows. Here, we present a benchmark experiment using two human lung adenocarcinoma cell lines that were each profiled in triplicate together with synthetic, spliced, spike-in RNAs (sequins). Samples were deeply sequenced on both Illumina short-read and Oxford Nanopore Technologies long-read platforms. Alongside the ground-truth available via the sequins, we created in silico mixture samples to allow performance assessment in the absence of true positives or true negatives. Our results show that StringTie2 and bambu outperformed other tools from the six isoform detection tools tested, DESeq2, edgeR and limma-voom were best among the five differential transcript expression tools tested and there was no clear front-runner for performing differential transcript usage analysis between the five tools compared, which suggests further methods development is needed for this application.

Systematic assessment of long-read RNA-seq methods for transcript identification and quantification

The Long-read RNA-Seq Genome Annotation Assessment Project (LRGASP) Consortium was formed to evaluate the effectiveness of long-read approaches for transcriptome analysis. The consortium generated over 427 million long-read sequences from cDNA and direct RNA datasets, encompassing human, mouse, and manatee species, using different protocols and sequencing platforms. These data were utilized by developers to address challenges in transcript isoform detection and quantification, as well as de novo transcript isoform identification. The study revealed that libraries with longer, more accurate sequences produce more accurate transcripts than those with increased read depth, whereas greater read depth improved quantification accuracy. In well-annotated genomes, tools based on reference sequences demonstrated the best performance. When aiming to detect rare and novel transcripts or when using reference-free approaches, incorporating additional orthogonal data and replicate samples are advised. This collaborative study offers a benchmark for current practices and provides direction for future method development in transcriptome analysis.

Challenges in identifying mRNA transcript starts and ends from long-read sequencing data

Long-read sequencing (LRS) technologies have the potential to revolutionize scientific discoveries in RNA biology, especially by enabling the comprehensive identification and quantification of full length mRNA isoforms. However, inherently high error rates make the analysis of long-read sequencing data challenging. While these error rates have been characterized for sequence and splice site identification, it is still unclear how accurately LRS reads represent transcript start and end sites. Here, we systematically assess the variability and accuracy of mRNA terminal ends identified by LRS reads across multiple sequencing platforms. We find substantial inconsistencies in both the start and end coordinates of LRS reads spanning a gene, such that LRS reads often fail to accurately recapitulate annotated or empirically derived terminal ends of mRNA molecules. To address this challenge, we introduce an approach to condition reads based on empirically derived terminal ends and identified a subset of reads that are more likely to represent full-length transcripts. Our approach can improve transcriptome analyses by enhancing the fidelity of transcript terminal end identification, but may result in lower power to quantify genes or discover novel isoforms. Thus, it is necessary to be cautious when selecting sequencing approaches and/or interpreting data from long-read RNA sequencing.