Biology subjects

Kudla, G.

Publications and source records attributed to Kudla, G..

At least 19 recordsLinked to original sources

bioRxiv2026

Mechanism-selective deep mutational scanning distinguishes ERCC2 disease phenotypes

Pathogenic ERCC2 variants cause xeroderma pigmentosum (XP), trichothiodystrophy (TTD) or both, yet variant effect scores are usually interpreted only as measures of pathogenicity rather than of which disease mechanism is disrupted. XPD, the ERCC2-encoded TFIIH subunit, functions in both nucleotide excision repair and transcription. Using yeast complementation deep mutational scanning, we measured the effects of nearly all XPD amino acid substitutions. The assay was mechanism-selective: it preferentially reported transcription-associated function, with pronounced intolerance at the p44 interface, whereas many substitutions affecting DNA binding and helicase activity retained near-wild-type fitness. Accordingly, TTD variants had much lower fitness than XP variants. Computational predictors discriminated pathogenic from benign variants similarly across phenotypes, but the DMS distinguished XP from TTD variants better than all 73 predictors tested. Phenotype-specific ACMG/AMP calibration provided evidence in both directions for TTD but mainly pathogenic evidence for XP. Thus, the selectivity of functional assays, often viewed as a limitation, can reveal disease mechanisms and support phenotype-aware variant interpretation.

genetics↗

bioRxiv2026

Evaluating the performance of splicing predictors on thousands of synthetic gene variants

Computational predictors of RNA splicing are increasingly used to interpret genetic variants and to design synthetic genes, yet they are almost always benchmarked on endogenous human sequences closely related to their training data. Whether their performance reflects genuine recognition of splicing signals, or instead exploits statistical features of natural genomes such as conservation and exon-intron composition, remains unclear. Here we benchmark eleven splicing predictors on thousands of synthetic GFP variants that are heavily recoded and dissimilar from any training data, using long-read sequencing to measure splicing directly at each position. Despite this distribution shift, modern deep-learning predictors retained strong performance, and the resulting ranking was largely stable across position-level and construct-level benchmarks. SpliceTransformer ranked highest, followed by AlphaGenome and SpliceAI. Tools that ignore long-range sequence context performed substantially worse, largely because they assign high scores to many non-spliced positions. This ranking broadly agrees with benchmarks on endogenous variants, indicating that the leading models capture transferable, sequence-intrinsic determinants of splicing. We further provide a unified calibration that maps each predictor's scores onto the measured fraction of spliced reads, allowing scores to be interpreted as splicing outcomes and compared directly between tools. Our results show that current deep-learning models generalise beyond natural genomes and provide a practical framework for splicing-aware sequence design.

genomics↗

bioRxiv2026

Splice-Aware Optimization Prevents Pervasive Missplicing of Natural and Synthetic cDNAs

Heterologous gene expression is widely used across biology and medicine, and often relies on codon optimization to increase protein yields. Here we uncover missplicing as a common and largely unrecognized failure mode of heterologous expression. Using systematically designed libraries comprising over 5,000 synthetic reporter genes and natural human cDNAs, we find that the majority of gene variants expressed in a human cell line are at least partially spliced, and in many variants the spliced isoform dominates, reducing protein output or ablating expression entirely. By analysing sequence determinants of expression across multiple human cell lines, we uncover a hierarchical architecture of regulatory control, where GC content establishes baseline mRNA levels, local sequence features influence splicing, and tissue-specific codon adaptation to tRNA pools fine-tunes translation efficiency. These findings enable us to develop predictive models of expression and splicing, benchmark current optimization strategies, and design a splice-aware optimization algorithm that substantially improves transgene performance.

synthetic biology↗

bioRxiv2026

Benchmarking siRNA Prediction: The Role of Representation and Validation Strategies

Small interfering RNAs (siRNAs) offer transformative potential for targeted therapeutics, yet the design of highly effective and non-toxic candidates is hindered by the risk of off-target effects and RNA instability. A critical flaw in in silico prediction models is pervasive data leakage in cross-validation protocols, which artificially inflates performance metrics and produces untrustworthy results. To address this, we developed a rigorous framework that eliminates data leakage through strict cross-validation, leverages z-curves (3D representations of RNA physico-chemical properties) for context-aware sequence encoding, and identifies key sequence regions critical for efficacy. Our model achieves an AUC of 0.845 on leakage-free validation, surpassing prior work at 380x faster computation speed, demonstrating that superior representation trumps model complexity. Crucially, we demonstrate how experimental variability and cross-validation choices directly impact model reliability, establishing the first benchmarked methods for robust siRNA efficacy prediction. This work provides a foundation for trustworthy sequence design and validation in RNA therapeutics.

bioinformatics↗

bioRxiv2026

CPLfold: Chimeric and Pseudoknot-capable almost Linear-time RNA Secondary Structure Prediction

Motivation RNA structure plays a central role in how transcripts function, but inferring it reliably remains difficult, especially when pseudoknots need to be part of the prediction. Chemical probing experiments provide additional signals, yet these signals do not directly identify base pairing partners. RNA proximity ligation provides direct evidence of base pairing, but balancing this evidence with pseudoknot prediction accuracy and scalability of structure prediction for long sequences remains challenging. Results We present CPLfold, a fast and flexible RNA folding method that combines thermodynamic modeling with chimeric evidence from RNA cross-linking and ligation experiments, while naturally supporting pseudoknots. CPLfold scales to long sequences and recovers more accurate global structures and long-range interactions than existing approaches across multiple benchmarks such as COMRADES and IRIS. By tuning two simple trade-off parameters (, {beta}) the method allows flexibility in the level of incorporating chimeric evidence and asserting pseudoknots. Availability and Implementation Source code and scripts are available at https://github.com/Vicky-0256/CPLfold. ContactK.Wang-72@sms.ed.ac.uk

bioinformatics↗

bioRxiv2026

A Functional Genetic Atlas of Parkin Resolves Variants of Uncertain Significance and Predicts Parkinson's Disease Age at Onset

Autosomal recessive mutations in the Parkin gene (PRKN) cause early-onset Parkinsons disease (PD). Parkin functions as a ubiquitin E3 ligase acting downstream of the PINK1 kinase to promote phosphorylated ubiquitin at sites of mitochondrial damage. Yet the functional effects of most PRKN gene variants remain unknown. Here we use a pooled cellular assay measuring phosphorylated ubiquitin accumulation to quantify the activity of [~]9,200 PRKN missense and nonsense variants. Our screen identifies thousands of loss- and gain-of-function variants, including activating variants mainly residing at autoinhibitory interfaces. Functional scores accurately distinguish known PD pathogenic and benign variants, and enable reclassification of 173/184 variants of uncertain significance (VUS). When combined into biallelic genotype scores, these data predict the age at disease onset in patients, revealing a quantitative link between Parkin activity and clinical manifestation. Our results provide a comprehensive functional genetic map of Parkin and demonstrate the power of multiplexed assays of variant effects (MAVEs) for variant interpretation and precision medicine in PD.

genetics↗

bioRxiv2026

3D confinement reshapes RNA folding and enhances circularisation in the Zika virus

Many RNA molecules function within confined environments, but the effect of confinement on RNA folding remains poorly understood. Proximity ligation experiments reveal altered long-range contacts in confined versus unconfined states, yet they do not explain how spatial constraints give rise to these differences. Here, we develop a physical modeling approach that incorporates proximity ligation data into coarse-grained molecular dynamics simulations to reconstruct RNA 3D structures under confinement. We test our model on the[~] 11 kb genome of the Zika virus, comparing the folding in virions (confined) and in cells (unconfined). We observe that the probability of contact between two regions of the genome vs. linear distance follows different scaling laws in the confined and unconfined cases, in agreement with proximity ligation experiments. We find that genome circularization - an interaction that regulates replication in Zika - occurs more frequently under confinement in both experiments and simulations. Our model reveals that the formation of long double-stranded stems through stacking confers local nematic liquid crystalline order to the RNA, and predicts pseudoknot topologies consistent with those seen in crystallographic structures of shorter RNAs. These results provide insight into the mechanism through which confinement alters the ensemble of 3D structures accessible to a long RNA molecule. Significance statementSqueezing a long string into a tight ball changes how it behaves: distant segments are forced into close contact, making ends more likely to meet. Similar principles apply to biopolymers, such as RNA, when confined within virus particles or other enclosures smaller than the polymer contour length, but there is no framework to model the effects of confinement on RNA folding. Here we develop such a framework, combining molecular dynamics simulations with experimental measurements of contacts within unconfined and confined RNA molecules. We find that RNA is glassy and exhibits liquid crystalline order, and that genome circularization, a long-range interaction between the two ends of the viral RNA important for virus replication, is facilitated under confinement.

biophysics↗

bioRxiv2025

Human Mendelian disease and in vivo mutagenesis screening define the molecular architecture of the U8 snoRNA

Over 50 mutations in the vertebrate specific box C/D small nucleolar RNA U8 have been reported to cause leukoencephalopathy with calcifications and cysts (LCC), a progressive cerebral microangiopathy presenting at any age between early infancy and late adulthood. Notably, the majority of these disease-associated mutations are annotated as variants of uncertain significance due to an inability to predict their effect on U8 function. Here, using a zebrafish bioassay, we performed in vivo mutagenesis screening to explore the molecular pathology of LCC. We demonstrate that LCC-associated mutations cluster in function and molecular behaviour according to distinct structural domains within U8, differentially impairing U8 processing, activity and stability. Further, we show that when U8 function is hypomorphic, exogenous human U8 and the endogenous zebrafish U8.3 paralogue increase in stability through a newly discovered post-transcriptional mechanism. This latter response is compromised in patients with LCC due to the presence of a common polymorphism of an N6-methyladenosine modified nucleotide, which we predict to impact disease penetrance.

neuroscience↗

bioRxiv2025

Optimization of regulatory DNA with active learning

Many biotechnology applications rely on microbial strains engineered to express heterologous proteins at maximal yield. A common strategy for improving protein output is to design expression systems with optimized regulatory DNA elements. Recent advances in high-throughput experimentation have enabled the use of machine learning predictors in tandem with sequence optimizers to find regulatory sequences with improved phenotypes. Yet the narrow coverage of training data, limited model generalization, and highly nonconvex nature of genotype-phenotype landscapes can limit the use of traditional sequence optimization algorithms. Here, we explore the use of active learning as a strategy to improve expression levels through iterative rounds of measurements, model training, and sequence sampling-and-selection. We explore convergence and performance of the active learning loop using synthetic data and an experimentally characterized genotype-phenotype landscape of yeast promoter sequences. Our results show that active learning can outperform one-shot optimization approaches in complex landscapes with a high degree of epistasis. We demonstrate the ability of active learning to effectively optimize sequences using datasets from different experimental conditions, with potential for leveraging data across laboratories, strains or growth conditions. Our findings highlight active learning as an effective framework for DNA sequence design, offering a powerful strategy for phenotype optimization in biotechnology.

synthetic biology↗

bioRxiv2024

Deep mutational scanning of the human insulin receptor ectodomain to inform precision therapy for insulin resistance

The insulin receptor (INSR) entrains tissue growth and metabolism to nutritional conditions. Complete loss of function in humans leads to extreme insulin resistance and infantile mortality, while loss of 80-90% function permits longevity of decades. Even low-level activation of severely compromised receptors, for example by anti-receptor monoclonal antibodies, thus offers the potential for decisive clinical benefit. A barrier to genetic diagnosis and translational research is the increasing identification of INSR variants of uncertain significance. We employed saturation mutagenesis coupled to multidimensional flow-based assays to stratify approximately 14,000 INSR extracellular domain missense variants by cell surface expression, insulin binding, and insulin- or monoclonal antibody-stimulated signaling. The resulting function scores correlate strongly with clinical syndromes, offer insights into dynamics of insulin binding, and reveal novel potential gain-of-function variants. This INSR sequence-function map has high biochemical, diagnostic and translational utility, aiding rapid identification of variants amenable to activation by non-canonical INSR agonists.

genetics↗

bioRxiv2024

Mapping HIV-1 RNA Structure, Homodimers, Long-Range Interactions and persistent domains by HiCapR

Human Immunodeficiency Virus (HIV) persists as a leading global health issue. A significant gap in knowledge exists in our understanding of long-range interactions within the HIV-1 RNA genome. To bridge this gap, we introduce HiCapR, incorporating a psoralen crosslinking RNA proximity ligation and post-library hybridization for capturing HIV RNA-RNA interactions. Leveraging HiCapR, we confirm the presence of stem structures in the key regions, such as the 5-untranslated region(5-UTR) and Rev Response Element (RRE) stems, and dimer sites in 5-UTR region, which is responsible for HIV packaging. Importantly, we reveal multiple previously unknown homodimers along the HIV genome, which may have important implications for viral RNA splicing and packaging processes. Also, we uncover a wealth of unprecedented long-range interactions, particularly within the 5-UTR of infected cells. Intriguingly, our findings indicate a pronounced reduction in long-range RNA-RNA interactions, signifying a transition from a state of abundant interactions, hence a relative loose state within infected cells to a condensed structure within virions. Concurrently, we have demonstrated the presence of stable genomic domains within virions that are instrumental in the dimerization process. These domains are preserved throughout the packaging process. Our findings shed light on the functional significance of RNA organization, including stable and persistent genomic domains, homodimerization, and long-range RNA-RNA interactions, in the splicing, packaging as well as assembly of HIV. HighlightsHiCapR is a new proximity ligation method for mapping RNA structures and homodimers in the HIV genome with sufficient reliability and efficiency. Multiple homodimers were discovered along the genome, with potential implications for splicing and packaging processes. Long-range RNA-RNA interactions are abundant in infected cells but significantly reduced in virions. Stable genomic domains encluding homodimer sites are persistent in virions and are involved in dimerization.

genomics↗

bioRxiv2024

DNA representations and generalization performance of sequence-to-expression models

The growing demand for biological products drives many efforts to maximize expression of heterologous proteins. Advances in high-throughput sequencing can produce data suitable for building sequence-to-expression models with machine learning. The most accurate models have been trained on one-hot encodings, a mechanism-agnostic representation of nucleotide sequences. Moreover, studies have consistently shown that training on mechanistic sequence features leads to much poorer predictions, even with features that are known to correlate with expression, such as DNA sequence motifs, codon usage or mRNA secondary structures. However, despite their excellent local accuracy, current sequence-to-expression models can fail to generalize predictions far away from the training data. Through a comparative study across datasets in Escherichia coli and Saccharomyces cerevisiae, here we show that mechanistic sequence features can provide gains on model generalization, and thus improving their utility for predictive sequence design. We explore several strategies to integrate one-hot encodings and mechanistic features into a single predictive model, including feature stacking, ensemble model stacking, and geometric stacking, a novel architecture based on graph convolutional neural networks. Our work casts new light on mechanistic sequence features, underscoring the importance of domain-knowledge and feature engineering for accurate prediction of protein expression levels.

synthetic biology↗

bioRxiv2023

Deep mutational scanning quantifies DNA binding and predicts clinical outcomes of PAX6 variants

Nonsense and missense mutations in the transcription factor PAX6 cause a wide range of eye development defects, including aniridia, microphthalmia and coloboma. To understand how changes of PAX6:DNA binding cause these phenotypes, we combined saturation mutagenesis of the paired domain of PAX6 with a yeast one-hybrid (Y1H) assay in which expression of a PAX6-GAL4 fusion gene drives antibiotic resistance. We quantified binding of more than 2,700 single amino-acid variants to two DNA sequence elements. Mutations in DNA-facing residues of the N-terminal subdomain and linker region were particularly detrimental, as were mutations to prolines and to negatively charged residues. Many variants caused molecular gain-of-function effects, including variants in position Ile71 that increased binding to the LE9 enhancer but decreased binding to a SELEX-derived binding site. Benchmarking against known patient variants and applying ACMG/AMP guidelines to variant classification, we obtained supporting to moderate evidence to suggest that 1,306 variants are likely benign, and 977, likely pathogenic. Our analysis shows that most pathogenic mutations in the paired domain of PAX6 can be explained simply by the effects of these mutations on PAX6:DNA association, and establishes Y1H as a generalisable assay for the interpretation of variant effects in transcription factors.

genetics↗

bioRxiv2022

The RNA-binding landscape of HAX1 protein indicates its involvement in ribosome biogenesis and translation.

HAX1 is a human protein with no known homologues or structural domains, mutations in which cause severe congenital neutropenia through mechanisms that are poorly understood. Previous studies reported RNA-binding capacity of HAX1, but the role of this binding in physiology and pathology remains unexplained. Here we report transcriptome-wide characterization of HAX1 RNA targets using RIP-seq and CRAC, indicating that HAX1 binds transcripts involved in ribosome biogenesis and rRNA processing. Using CRISPR knockouts we find that RNA targets of HAX1 partially overlap with transcripts downregulated in HAX1 KO, implying a role in mRNA stabilization. Gene ontology analysis demonstrated that genes differentially expressed in HAX1 KO (including genes involved in ribosome biogenesis and translation) are also enriched in a subset of genes whose expression correlates with HAX1 expression in four analyzed neoplasms. Functional connection to ribosome biogenesis was also demonstrated by gradient sedimentation ribosome profiles, which revealed differences in the small subunit:monosome ratio in HAX1 WT/KO. We speculate that changes in HAX1 expression may be important for the etiology of HAX1-linked diseases through dysregulation of translation.

molecular biology↗

bioRxiv2021

Structural features within the NORAD long noncoding RNA underlie efficient repression of Pumilio activity

It is increasingly appreciated that long non-coding RNAs (lncRNAs) carry out important functions in mammalian cells, but how these are encoded in their sequences and manifested in their structures remains largely unknown. Some lncRNAs bind to and modulate the availability of RNA binding proteins, but the structural principles that underlie this mode of regulation are underexplored. Here, we focused on the NORAD lncRNA, which binds Pumilio proteins and modulates their ability to repress hundreds of mRNA targets. We probed the RNA structure and long-range RNA-RNA interactions formed by NORAD inside cells, under different stressful conditions. We discovered that NORAD structure is highly modular, and consists of well-defined domains that contribute independently to NORAD function. We discovered that NORAD structure spatially clusters the Pumilio binding sites along NORAD in a manner that contributes to the de-repression of Pumilio target proteins. Following arsenite stress, the majority of NORAD structure undergoes relaxation and forms inter-molecular interactions with RNAs that are targeted to stress granules. NORAD sequence thus dictates elaborated structural domain organization that facilitates its function on multiple levels, and which helps explain the extensive evolutionary sequence conservation of NORAD regions that are not predicted to directly bind Pumilio proteins.

molecular biology↗

bioRxiv2021

Global mapping of RNA homodimers in living cells

RNA homodimerization is important for various physiological processes, including the assembly of membraneless organelles, RNA subcellular localization, and packaging of viral genomes. However, understanding of RNA homodimerization has been hampered by the lack of systematic in vivo detection methods. Here we show that PARIS, COMRADES, and other RNA proximity ligation methods can detect RNA homodimers transcriptome-wide as "overlapping" chimeric reads that contain more than one copy of the same sequence. Analysing published proximity ligation datasets, we show that RNA:RNA homodimers mediated by direct base-pairing interactions are rare across the transcriptome, but highly enriched in specific transcripts, including U8 snoRNA, U2 snRNA and a subset of tRNAs. Analysis of data from virus-infected cells reveals homodimerization of SARS-CoV-2 and Zika genomes, mediated by specific palindromic sequences located within protein-coding regions of N protein in SARS-CoV-2 and NS2A gene in Zika. We speculate that regions of viral genomes involved in homodimerization may constitute effective targets for antiviral therapies.

systems biology↗

bioRxiv2021

In vivo structure and dynamics of the RNA genome of SARS-Cov-2

The SARS-CoV-2 coronavirus, which causes the COVID-19 pandemic, is one of the largest positive strand RNA viruses. Here we developed a simplified SPLASH assay and comprehensively mapped the in vivo RNA-RNA interactome of SARS-CoV-2 RNA during the viral life cycle. We observed canonical and alternative structures including 3-UTR and 5-UTR, frameshifting element (FSE) pseudoknot and genome cyclization in cells and in virions. We provide direct evidence of interactions between Transcription Regulating Sequences (TRS-L and TRS-Bs), which facilitate discontinuous transcription. In addition, we reveal alternative short and long distance arches around FSE, forming a "high-order pseudoknot" embedding FSE, which might help ribosome stalling at frameshift sites. More importantly, we found that within virions, while SARS-CoV-2 genome RNA undergoes intensive compaction, genome cyclization is weakened and genome domains remain stable. Our data provides a structural basis for the regulation of replication, discontinuous transcription and translational frameshifting, describes dynamics of RNA structures during life cycle of SARS-CoV-2, and will help to develop antiviral strategies.

genomics↗

bioRxiv2020

Genotype-phenotype map of an RNA-ligand complex

RNA-ligand interactions play important roles in biology and biotechnology, but they often involve complex three-dimensional folding of RNA and are difficult to predict. To systematically explore the phenotypic landscape of an RNA-ligand complex, we used microarrays to investigate all possible single and double mutants of the 49-nt RNA aptamer Broccoli bound to the fluorophore DFHBI-1T. We collected more than seven million fluorescence measurements in varying conditions, and inferred dissociation rate constants, spectral shifts, and intragenic epistasis. Our results reveal an unexpectedly complex phenotypic landscape, in which mutations near the fluorophore binding pocket modulated magnesium-, potassium- and fluorophore-binding and fluorescence spectra, while distal mutations influenced structural stability and fluorescence intensity. We trained a machine learning model that accurately predicted RNA secondary structure from local epistatic interactions, despite the presence of G-quadruplexes and other noncanonical structures. Our experimental platform will facilitate the discovery and analysis of new RNA-ligand interactions.

synthetic biology↗