Biology subjects

Moyon, L.

Publications and source records attributed to Moyon, L..

9 recordsLinked to original sources

bioRxiv2026

PARNET: A CLIP-SEQ-BASED FOUNDATION MODEL FOR RNA SEQUENCE REPRESENTATION LEARNING

RNA-binding proteins (RBPs) orchestrate a complex combinatorial regulatory "code" that governs RNA splicing, stability, localization, and translation. Learning the relationship between RNA sequences and these processes is a central challenge in genomics. Foundation models, notably RNA language models, have emerged as the dominant approach, learning general-purpose representations from unlabeled sequence at scale. While RNA language models have demonstrated impressive performance across a broad range of downstream tasks, they generally learn from sequence reconstruction objectives alone, lacking direct connections to the regulatory principles that govern RNA function. Here we introduce Parnet, an RNA foundation model trained directly and exclusively on experimental CLIP-seq data. Parnet is a multi-task foundation model trained end-to-end on 223 eCLIP-seq experiments spanning 150 RBPs to predict base-resolution RBP binding profiles directly from RNA sequence. This CLIP-seq pretraining strategy departs fundamentally from the masked-language-modeling paradigm, anchoring learned RNA representations directly in measured protein-RNA interactions rather than sequence statistics. Parnet substantially outperforms its single-task predecessor RBPNet in binding profile and motif recovery, generalizes to unseen cell types and iCLIP data, and recapitulates position-dependent splicing regulation. Frozen Parnet embeddings, without task-specific fine-tuning, match or exceed the performance of both task-specific tools, as well as larger self-supervised RNA and genomic language models across diverse downstream tasks, including RNA biotype classification, lncRNA chromatin localization, translational efficiency, splice-site recognition, intron retention, and non-coding variant effect prediction. Importantly, Parnet remains mechanistically interpretable, tracing predictions back to the specific RBPs and motifs that drive them. These results establish the RBP interactome as a compact, functionally sufficient, and interpretable basis for foundation model pretraining in RNA biology.

bioinformatics↗

bioRxiv2026

Pan-cancer discovery of driver mutations in long noncoding RNAs reveals widespread functional rewiring of RNA regulatory elements

Most somatic mutations in cancer occur outside protein-coding genes, yet the functional impact of these mutations remains largely unknown. Long noncoding RNAs (lncRNAs) represent a major class of cancer-promoting genes whose molecular mechanisms are poorly understood. While individual driver mutations in lncRNAs have been identified, detecting such driver lncRNAs at scale requires large tumour genome cohorts. We analyse 12,631 cancer genomes from the 100,000 Genomes Project (100kGP) and identify 121 lncRNAs under positive selection across 19 cancer types. These driver lncRNAs are independently supported by functional genomic screens, germline predisposing variants, mutual exclusivity with protein-coding drivers, and independent oncogenic lncRNA catalogues. Overall, approximately two-thirds of analysed tumours harbour at least one lncRNA driver mutation. Leveraging the depth of this dataset, we demonstrate that somatic mutations preferentially target and remodel RNA-binding protein (RBP) interaction sites to potentiate oncogenic lncRNAs, including MALAT1, SNHG14 and NEAT1. From these data, we derive a model in which somatic mutations liberate oncogenic lncRNAs from repressive RNA:protein interactions. This work expands the number and nature of cancer driver genes, identifies targets for RNA-directed therapies, and demonstrates that with large tumour mutation catalogues we can dissect the molecular mechanisms of noncoding genes.

Cancer Biology↗

bioRxiv2025

Integrative Gene and Isoform Co-expression Networks Reveal Regulatory Rewiring in Stress-related Psychiatric Disorders

Isoform-specific expression patterns have been implicated in stress-related psychiatric disorders like major depressive disorder (MDD), yet the extent of their involvement and their interrelationships remain unclear. We constructed co-expression networks for individuals affected (n=210, 81% with depressive symptoms) and unaffected (n=95) by stress-related psychiatric disorders. We incorporated total gene expression (TE) and isoform ratio (IR) data and validated the inferred networks using advanced graph generation techniques. Our analysis revealed distinct network topology and structure between the two groups. Investigation of the 127 shared hubs (degree >= 10) found that these hubs exhibit co-regulatory patterns unique to each network. The affected individuals network also contained 61 hub nodes with a minimum absolute fold increase of two in connectivity compared to the unaffected individuals network. Notably, 49% of these hubs showed evidence for association with psychiatric disorders. Gene Ontology enrichment analysis revealed distinct biological processes associated with hubs, such as mRNA processing for affected and immune response and cell adhesion for unaffected individuals. Enrichment analysis of GWAS loci further supported network-specific findings. Analysis of the isoform-specific nodes showed distinct protein-protein interactions compared to gene-level analysis. This is the first study to demonstrate network-level differences in gene and isoform co-expression patterns between individuals with and without stress-related psychiatric disorders, with a particular focus on depressive symptoms. Our findings provide evidence for substantial rewiring of gene regulatory networks in affected individuals. Incorporating isoform-level data revealed a deeper level of complexity, highlighting the importance of considering isoform variations in understanding the molecular basis of these conditions.

systems biology↗

bioRxiv2023

Combinatorial microRNA activity is essential for the transition of pluripotent cells from proliferation into dormancy

Dormancy is a key feature of stem cell function in adult tissues as well as embryonic cells in the context of diapause. The establishment of dormancy is an active process that involves extensive transcriptional, epigenetic, and metabolic rewiring1,2. How these processes are coordinated to successfully transition cells to the resting dormant state remains unclear. Here we show that microRNA activity, which is otherwise dispensable for pre-implantation development, is essential for the adaptation of early mouse embryos to the dormant state of diapause. In particular, the pluripotent epiblast depends on miRNA activity, the absence of which results in loss of pluripotent cells. Through integration of high-sensitivity small RNA expression profiling of individual embryos and protein expression of miRNA targets with public data of protein-protein interactions, we constructed the miRNA-mediated regulatory network of mouse early embryos specific to diapause. We find that individual miRNAs contribute to the combinatorial regulation by the network and the perturbation of the network compromises embryo survival in diapause. Without miRNAs, nuclear and cytoplasmic bodies show aberrant expression, concurrent with splicing defects. We identified the nutrient-sensitive transcription factor TFE3 as an upstream regulator of diapause-specific miRNAs, linking cytoplasmic mTOR activity to nuclear miRNA biogenesis. Our results place miRNAs as a critical regulatory layer for the molecular rewiring of early embryos to establish dormancy.

molecular biology↗

bioRxiv2023

A Systematic Benchmark of Machine Learning Methods for Protein-RNA Interaction Prediction

AO_SCPLOWBSTRACTC_SCPLOWRNA-binding proteins (RBPs) are central actors of RNA post-transcriptional regulation. Experiments to profile binding sites of RBPs in vivo are limited to transcripts expressed in the experimental cell type, creating the need for computational methods to infer missing binding information. While numerous machine-learning based methods have been developed for this task, their use of heterogeneous training and evaluation datasets across different sets of RBPs and CLIP-seq protocols makes a direct comparison of their performance difficult. Here, we compile a set of 37 machine learning (primarily deep learning) methods for in vivo RBP-RNA interaction prediction and systematically benchmark a subset of 11 representative methods across hundreds of CLIP-seq datasets and RBPs. Using homogenized sample pre-processing and two negative-class sample generation strategies, we evaluate methods in terms of predictive performance and assess the impact of neural network architectures and input modalities on model performance. We believe that this study will not only enable researchers to choose the optimal prediction method for their tasks at hand, but also aid method developers in developing novel, high-performing methods by introducing a standardized framework for their evaluation.

bioinformatics↗

bioRxiv2022

Towards In-Silico CLIP-seq: Predicting Protein-RNA Interaction via Sequence-to-Signal Learning

AO_SCPLOWBSTRACTC_SCPLOWUnraveling sequence determinants which drive protein-RNA interaction is crucial for studying binding mechanisms and the impact of genomic variants. While CLIP-seq allows for transcriptome-wide profiling of in vivo protein-RNA interactions, it is limited to expressed transcripts, requiring computational imputation of missing binding information. Existing classification-based methods predict binding with low resolution and depend on prior labeling of transcriptome regions for training. We present RBPNet, a novel deep learning method, which predicts CLIP crosslink count distribution from RNA sequence at single-nucleotide resolution. By training on up to a million regions, RBPNet achieves high generalization on eCLIP, iCLIP and miCLIP assays, outperforming state-of-the-art classifiers. CLIP-seq suffers from various technical biases, complicating downstream interpretation. RBPNet performs bias correction by modeling the raw signal as a mixture of the protein-specific and background signal. Through model interrogation via Integrated Gradients, RBPNet identifies predictive sub-sequences corresponding to known binding motifs and enables variant-impact scoring via in silico mutagenesis. Together, RBPNet improves inference of protein-RNA interaction, as well as mechanistic interpretation of predictions.

bioinformatics↗

bioRxiv2021

Computational Mapping of the Human-SARS-CoV-2 Protein-RNA Interactome

Strong evidence suggests that human human RNA-binding proteins (RBPs) are critical factors for viral infection, yet there is no feasible experimental approach to map exact binding sites of RBPs across the SARS-CoV-2 genome systematically at a large scale. We investigated the role of RBPs in the context of SARS-CoV-2 by constructing the first in silico map of human RBP / viral RNA interactions at nucleotide-resolution using two deep learning methods (pysster and DeepRiPe) trained on data from CLIP-seq experiments. We evaluated conservation of RBP binding between 6 other human pathogenic coronaviruses and identified sites of conserved and differential binding in the UTRs of SARS-CoV-1, SARS-CoV-2 and MERS. We scored the impact of variants from 11 viral strains on protein-RNA interaction, identifying a set of gain-and loss of binding events. Lastly, we linked RBPs to functional data and OMICs from other studies, and identified MBNL1, FTO and FXR2 as potential clinical biomarkers. Our results contribute towards a deeper understanding of how viruses hijack host cellular pathways and are available through a comprehensive online resource (https://sc2rbpmap.helmholtz-muenchen.de).

bioinformatics↗

bioRxiv2021

Classification of non-coding variants with high pathogenic impact

Whole genome sequencing is increasingly used to diagnose medical conditions of genetic origin. While both coding and non-coding DNA variants contribute to a wide range of diseases, most patients who receive a WGS-based diagnosis today harbour a protein-coding mutation. Functional interpretation and prioritization of non-coding variants represents a persistent challenge, and disease-causing non-coding variants remain largely unidentified. Depending on the disease, WGS fails to identify a candidate variant in 20-80% of patients, severely limiting the usefulness of sequencing for personalised medicine. Here we present FINSURF, a machine-learning approach to predict the functional impact of non-coding variants in regulatory regions. FINSURF outperforms state-of-the-art methods, owing to control optimisation during training. In addition to ranking candidate variants, FINSURF also delivers diagnostic information on functional consequences of mutations. We applied FINSURF to a diverse set of 30 diseases with described causative non-coding mutations, and correctly identified the disease-causative non-coding variant within the ten top hits in 22 cases. FINSURF is implemented as an online server to as well as custom browser tracks, and provides a quick and efficient solution to prioritize candidate non-coding variants in realistic clinical settings.

bioinformatics↗

bioRxiv2020

Functional Predictors of Causative Cis-Regulatory Mutations in Mendelian Disease

Undiagnosed neurodevelopmental disease is significantly associated with rare variants in cis-regulatory elements (CRE) but demonstrating causality is challenging as target gene consequences may differ from a causative variant affecting the coding region. Here, we address this challenge by applying a procedure to discriminate likely diagnostic regulatory variants from those of neutral/low-penetrant effect. We identified six rare CRE variants using targeted and whole genome sequencing in 48 unrelated males with apparent X-linked intellectual disability (XLID) but without detectable coding region variants. These variants segregated appropriately in families and altered conserved bases in predicted CRE targeting known XLID genes. Three were unique and three were rare but too common to be plausibly causative for XLID. We compared the cis-regulatory activity of wild-type and mutant alleles in zebrafish embryos using dual-color fluorescent reporters. Two variants showed striking changes: one plausibly causative (FMR1CRE) and the other likely neutral/low-penetrant (TENM1CRE). These variants were "knocked-in" to mice and both altered embryonic neural expression of their target gene. Only Fmr1CRE mice showed disease-relevant behavioral defects. FMR1CRE is plausibly disease-associated resulting in complex misregulation of Fmr1/FMRP rather than loss-of-function. This is consistent both with absence of Fragile X syndrome in the probands and the observed electrophysiological anomalies in the FMR1CRE mouse brain. Although disruption of in vivo patterns of endogenous gene expression in disease-relevant tissues by CRE variants cannot be used as strong evidence for Mendelian disease association, in conjunction with extreme rarity in human populations and with relevant knock-in mouse phenotypes, such variants can become likely pathogenic.

genetics↗