Deep learning and direct sequencing of labeled RNA captures transcriptome dynamics.

In eukaryotes, genes produce a variety of distinct RNA isoforms, each with potentially unique protein products, coding potential or regulatory signals such as poly(A) tail and nucleotide modifications. Assessing the kinetics of RNA isoform metabolism, such as transcription and decay rates, is essential for unraveling gene regulation. However, it is currently impeded by lack of methods that can differentiate between individual isoforms. Here, we introduce RNAkinet, a deep convolutional and recurrent neural network, to detect nascent RNA molecules following metabolic labeling with the nucleoside analog 5-ethynyl uridine and long-read, direct RNA sequencing with nanopores. RNAkinet processes electrical signals from nanopore sequencing directly and distinguishes nascent from pre-existing RNA molecules. Our results show that RNAkinet prediction performance generalizes in various cell types and organisms and can be used to quantify RNA isoform half-lives. RNAkinet is expected to enable the identification of the kinetic parameters of RNA isoforms and to facilitate studies of RNA metabolism and the regulatory elements that influence it.

Full-length direct RNA sequencing uncovers stress-granule dependent RNA decay upon cellular stress.

Cells react to stress by triggering response pathways, leading to extensive alterations in the transcriptome to restore cellular homeostasis. The role of RNA metabolism in shaping the cellular response to stress is vital, yet the global changes in RNA stability under these conditions remain unclear. In this work, we employ direct RNA sequencing with nanopores, enhanced by 5' end adaptor ligation, to comprehensively interrogate the human transcriptome at single-molecule and nucleotide resolution. By developing a statistical framework to identify robust RNA length variations in nanopore data, we find that cellular stress induces prevalent 5' end RNA decay that is coupled to translation and ribosome occupancy. Unlike typical RNA decay models in normal conditions, we show that stress-induced RNA decay is dependent on XRN1 but does not depend on removal of the poly(A) tail. We observed that RNAs undergoing decay are predominantly enriched in the stress granule transcriptome. Inhibition of stress granule formation via genetic ablation of G3BP1 and G3BP2 fully rescues RNA length and suppresses stress-induced decay. Our findings reveal RNA decay as a key determinant of RNA metabolism upon cellular stress and dependent on stress-granule formation.

Deep learning and direct sequencing of labeled RNA captures transcriptome dynamics.

Quantification of the dynamics of RNA metabolism is essential for understanding gene regulation in health and disease. Existing methods rely on metabolic labeling of nascent RNAs and physical separation or inference of labeling through PCR-generated mutations, followed by short-read sequencing. However, these methods are limited in their ability to identify transient decay intermediates or co-analyze RNA decay with cis-regulatory elements of RNA stability such as poly(A) tail length and modification status, at single molecule resolution. Here we use 5-ethynyl uridine (5EU) to label nascent RNA followed by direct RNA sequencing with nanopores. We developed RNAkinet, a deep convolutional and recurrent neural network that processes the electrical signal produced by nanopore sequencing to identify 5EU-labeled nascent RNA molecules. RNAkinet demonstrates generalizability to distinct cell types and organisms and reproducibly quantifies RNA kinetic parameters allowing the combined interrogation of RNA metabolism and cis-acting RNA regulatory elements.

Genomic benchmarks: a collection of datasets for genomic sequence classification.

BACKGROUND: Recently, deep neural networks have been successfully applied in many biological fields. In 2020, a deep learning model AlphaFold won the protein folding competition with predicted structures within the error tolerance of experimental methods. However, this solution to the most prominent bioinformatic challenge of the past 50 years has been possible only thanks to a carefully curated benchmark of experimentally predicted protein structures. In Genomics, we have similar challenges (annotation of genomes and identification of functional elements) but currently, we lack benchmarks similar to protein folding competition. RESULTS: Here we present a collection of curated and easily accessible sequence classification datasets in the field of genomics. The proposed collection is based on a combination of novel datasets constructed from the mining of publicly available databases and existing datasets obtained from published articles. The collection currently contains nine datasets that focus on regulatory elements (promoters, enhancers, open chromatin region) from three model organisms: human, mouse, and roundworm. A simple convolution neural network is also included in a repository and can be used as a baseline model. Benchmarks and the baseline model are distributed as the Python package 'genomic-benchmarks', and the code is available at https://github.com/ML-Bioinfo-CEITEC/genomic_benchmarks . CONCLUSIONS: Deep learning techniques revolutionized many biological fields but mainly thanks to the carefully curated benchmarks. For the field of Genomics, we propose a collection of benchmark datasets for the classification of genomic sequences with an interface for the most commonly used deep learning libraries, implementation of the simple neural network and a training framework that can be used as a starting point for future research. The main aim of this effort is to create a repository for shared datasets that will make machine learning for genomics more comparable and reproducible while reducing the overhead of researchers who want to enter the field, leading to healthy competition and new discoveries.