Biology subjects

Marsh, J. A.

Publications and source records attributed to Marsh, J. A..

At least 37 records · Page 2Linked to original sources

bioRxiv2025

Assessing variant effect predictors and disease mechanisms in intrinsically disordered proteins

Intrinsically disordered protein regions (IDPRs) are central to diverse cellular processes but present unique challenges for interpreting genetic variants implicated in human disease. Unlike structured protein domains, IDPRs lack stable three-dimensional conformations and are often involved in regulation through transient interactions and post-translational modifications. These features can affect both the distribution of pathogenic variants and the performance of computational tools used to predict their effects. Here, we systematically assessed the distribution of pathogenic vs benign missense variants across disordered, intermediate, and structured protein regions in the human proteome. Pathogenic variants were notably depleted in IDPRs yet were associated with distinct molecular mechanisms, particularly dominant gain- and loss-of-function effects. We evaluated 33 variant effect predictors (VEPs), revealing widespread reductions in sensitivity for pathogenic variants in IDPRs--despite high AUROC scores largely driven by accurate benign variant predictions. We also observed substantial discordance among VEP classifications in disordered regions, underscoring the need for region-aware thresholds and disorder-informed prediction strategies. Incorporating features reflective of IDPR biology, such as transient interaction motifs and modification sites, may enhance the accuracy and interpretability of future tools.

bioinformatics↗

bioRxiv2025

A structure-guided approach to non-coding variant evaluation for transcription factor binding using AlphaFold 3

Non-coding single-nucleotide variants (SNVs) that alter transcription factor (TF) binding can affect gene expression and contribute to disease. Sequence-based methods can excel at predicting TF binding, but rely on training data and can exhibit TF-specific biases. Here we propose a structure-guided approach for non-coding SNVs, using AlphaFold 3 (AF3) to model TF-DNA complexes and FoldX for downstream physics-based assessment. Benchmarked against SNP-SELEX data for six TFs (SPIB, ELK3, ETV4, SF-1, PAX5 and MEIS2), the FoldX-based strategy showed good agreement with experimental allele preferences. Interestingly, differences in AF3s interface predicted template modelling (ipTM) score aligned even more closely with SNP-SELEX results, generally surpassing energy-based metrics. Application to known disease-associated variants recapitulated most reported effects for TFs including NKX2-5, GATA3 and USF2A-USF1. In these examples, considering both {Delta}ipTM and FoldX energies proved more reliable than either metric alone. While less accurate than state-of-the-art sequence-based methods, this work demonstrates that structural modelling can yield interpretable insights into how non-coding variants influence TF binding. By highlighting both the promise and limitations of AF3 in this context, our study provides a framework for complementary structural evaluation of regulatory variants.

bioinformatics↗

bioRxiv2025

Prevalence of loss-of-function, gain-of-function and dominant-negative mechanisms across genetic disease phenotypes

Molecular disease mechanisms caused by mutations in protein-coding regions are diverse, but they can be broadly categorised into loss-of-function (LOF), gain-of-function (GOF), and dominant-negative (DN) effects. Accurately predicting these mechanisms is a pressing clinical need, as therapeutic strategies must align with the underlying disease mechanism. Moreover, computational predictors tend to perform less well at the identification of pathogenic GOF and DN variants. Here, we develop a protein structure-based missense LOF (mLOF) likelihood score that can separate recessive LOF and dominant LOF from alternative disease mechanisms. Using mLOF scores, we estimated the prevalence of molecular mechanisms across 2,837 phenotypes in 1,979 Mendelian disease genes, finding that DN and GOF mechanisms account for 48% of phenotypes in dominant genes. Applying mLOF scores to genes with multiple phenotypes revealed widespread intragenic mechanistic heterogeneity, with 43% of dominant and 49% of mixed-inheritance genes harbouring both LOF and non-LOF mechanisms. Furthermore, we show that combining mLOF scores with phenotype semantic similarity enables the prioritisation of DN mechanisms in mixed-inheritance genes. Our structure-based approach, accessible via a Google Colab notebook, offers a scalable tool for predicting disease mechanisms and advancing personalised medicine.

bioinformatics↗

bioRxiv2024

Structural extension of the human exocyst is enabled by a minimal interface

In multicellular organisms, the machinery responsible for polarized trafficking directs constitutive cargo secretion at distinct sites of the plasma membrane, cilia, and junctional structures. Central to this machinery is the exocyst complex, which tethers cargo vesicles to their destination membrane, alongside other intracellular membrane tethering roles. Precisely how the exocyst spatially integrates membranes and membrane resident binding partners is unclear. Here, we address the structural morphology and formation of the human exocyst complex. Through structural approaches coupled to predictive models, we determined that the exocyst and its subcomplexes have extended arm-like structures that help maximize its reach. Moreover, we demonstrate minimal intersubunit interaction, in contrast to prior models. Nucleation of the holocomplex occurs through a single site, explaining its spatial extension. Our results provide the biochemical basis for exocyst complex assembly, suggesting an ornate extended architecture. SignificanceCargo transport to the eukaryotic cell plasma membrane predominantly relies on the exocyst as a central, signal-integrating polarized trafficking complex. How this single cargo vesicle tethering complex can target diverse vesicles to destination membrane is poorly understood, partly due to inconsistent structural models. Here, we show an extended architecture of the human exocyst complex, revealing a minimal nucleation interface between its two subcomplexes. The distinct morphology of human exocyst, compared to yeast models, suggests a novel mechanism that supports its versatile role in membrane trafficking.

biochemistry↗

bioRxiv2024

Understanding the heterogeneous performance of variant effect predictors across human protein-coding genes

Variant effect predictors (VEPs) are computational tools developed to assess the impacts of genetic mutations, often in terms of likely pathogenicity, employing diverse algorithms and training data. Here, we investigate the performance of 35 VEPs in the discrimination between pathogenic and putatively benign missense variants across 963 human protein-coding genes, revealing considerable gene-level heterogeneity as measured by the widely used area under the receiver operating characteristic curve (AUROC) metric. To investigate the origins of this heterogeneity and the extent to which gene-level VEP performance is predictable, we train random forest models to predict the gene-level AUROC for each VEP. We find that performance as measured by AUROC is related to factors such as gene function, protein structure, and evolutionary conservation. Notably, intrinsic disorder in proteins emerged as a significant factor influencing apparent VEP performance, often leading to inflated AUROC values due to their enrichment in weakly conserved putatively benign variants. While our results suggest that gene-level features may be useful for identifying genes where VEP predictions are likely to be more or less reliable, they also highlight the limitations of AUROC for comparing VEP performance across different genes.

bioinformatics↗

bioRxiv2024

Structural Insight into the Function of Human Peptidyl Arginine Deiminase 6

Peptidyl arginine deiminase 6 (PADI6) is vital for early embryonic development in mice and humans, yet its function remains elusive. PADI6 is less conserved than other PADIs and it is currently unknown whether it has a catalytic function. Here we have shown that human PADI6 dimerises like hPADIs 2-4, however, does not bind Ca2+ and is inactive in in vitro assays against standard PADI substrates. By determining the crystal structure of hPADI6, we show that hPADI6 is structured in the absence of Ca2+ where hPADI2 and hPADI4 are not, and the Ca-binding sites are not conserved. Moreover, we show that whilst the key catalytic aspartic acid and histidine residues are structurally conserved, the cysteine is displaced far from the active site centre and the hPADI6 active site pocket appears closed through a unique evolved mechanism in hPADI6, not present in the other PADIs. Taken together, these findings provide insight into how the function of hPADI6 may differ from the other PADIs based on its structure and provides a resource for characterising the damaging effect of clinically significant PADI6 variants.

biochemistry↗

bioRxiv2024

Pervasive ancestry bias in variant effect predictors

Variant effect predictors (VEPs) - computational tools that assess the potential impact of genetic variants - have become increasingly vital for clinical variant interpretation. Currently, most VEPs used in variant prioritisation have been trained on datasets of clinically curated or population-derived variants. These datasets, however, disproportionately represent individuals of European descent. We hypothesised that this bias may lead to unequal VEP performance across different populations. To test this, we evaluated the scoring patterns of 52 VEPs for missense variants across 14 ancestry groups. We observe striking disparities: some VEPs predict a markedly higher proportion of damaging variants in underrepresented populations, such as those of Malay descent, compared to individuals of European ancestry. In contrast, VEPs that do not rely on clinical or population data predict more consistent pathogenicity burdens across ancestry groups. Moreover, we could closely link these discrepancies across methods to biases in training data. Our findings underscore the urgent need to adopt tools that minimise ancestry bias to ensure fairer and more accurate variant effect prediction and genetic diagnoses for all populations.

bioinformatics↗

bioRxiv2024

Variant effect predictor correlation with functional assays is reflective of clinical classification performance

BackgroundUnderstanding the relationship between protein sequence and function is crucial for accurate genetic variant classification. Variant effect predictors (VEPs) play a vital role in deciphering this complex relationship, yet evaluating their performance remains challenging for several reasons including data circularity, where the same or related data is used for training and assessment. High-throughput experimental strategies like deep mutational scanning (DMS) offer a promising solution. ResultsIn this study, we extend upon our previous benchmarking approach, assessing the performance of 97 different VEPs using DMS experiments from 36 different human proteins. In addition, a new pairwise, VEP-centric ranking method reduces the impact of missing predictions on the overall ranking. We observe a remarkably high correspondence between VEP performance in DMS-based benchmarks and clinical variant classification, especially for predictors that have not been directly trained on human clinical variants. ConclusionsOur results suggest that comparing VEP performance against diverse functional assays represents a reliable strategy for assessing their relative performance in clinical variant classification. However, major challenges in clinical interpretation of VEP scores persist, highlighting the need for further research to fully leverage computational predictors for genetic diagnosis. We also address practical considerations for end users in terms of choice of methodology.

bioinformatics↗

bioRxiv2024

Protein structural context of cancer mutations reveals molecular mechanisms and identifies novel candidate driver genes

Advances in structure determination and computational modelling are enabling us to study the protein structural context of human genetic variants at an unprecedented scale. Here, we investigate millions of human cancer-associated missense mutations in terms of their structural locations and predicted perturbative effects. We find that, while cancer-driving mutations have properties similar to other known disease-causing mutations, this is obscured by the abundance of passenger mutations in cancer sequencing datasets. Nevertheless, by considering the collective properties of mutations at the level of individual proteins, we identify distinct mutational signatures associated with tumour suppressors and oncogenes. Tumour suppressors are enriched in structurally damaging mutations, consistent with loss-of-function mechanisms. In contrast, oncogene mutations tend to be structurally mild, reflecting selection for gain-of-function driver mutations and against loss-of-function mutations. Although oncogenes are difficult to distinguish from genes with no role in cancer using only structural damage, we find that an alternate metric based on the clustering of mutations in three-dimensional space is highly predictive of oncogenes, particularly when mutation recurrence is considered. These observations allow us to identify novel candidate driver genes and speculate about their molecular roles, which we expect to have general utility in the analysis of cancer sequencing data.

bioinformatics↗

bioRxiv2023

Mutational scanning reveals oncogenic CTNNB1 mutations have diverse effects on signalling and clinical traits

CTNNB1, the gene encoding {beta}-catenin, is a frequent target for oncogenic mutations activating the canonical Wnt signalling pathway, typically via missense mutations within a degron hotspot motif in exon 3. Here, we combine saturation genome editing with a fluorescent reporter assay to quantify signalling phenotypes for all 342 missense mutations in the mutation hotspot, including 74 recurrent mutations reported in over 6000 tumours. Our data define the genetic requirements for {beta}-catenin degron function and reveal diverse levels of signal activation among known driver mutations. Tumorigenesis in different human tissues involves selection for CTNNB1 mutations spanning distinct ranges of effect size. In hepatocellular carcinoma, mutations that activate {beta}-catenin relatively weakly are associated with worse prognosis compared to stronger activating mutations, despite greater immune cell infiltration in the tumour microenvironment. Our work therefore provides a resource to understand mutational diversity within a pan-cancer mutation hotspot, with potential implications for targeted therapy.

cancer biology↗

bioRxiv2023

ALDH1A3-acetaldehyde re-wires neural crest stem cell and high metabolism states to potentiate melanoma heterogeneity

Cancer cellular heterogeneity and therapy resistance arise substantially from metabolic and transcriptional adaptations, but how these are interconnected is poorly understood. Here, we show that in melanoma, the cancer stem cell marker aldehyde dehydrogenase 1A3 (ALDH1A3) forms an enzymatic partnership with acetyl-CoA synthetase 2 (ACSS2) in the nucleus to couple high glucose metabolic flux with acetyl-histone H3 modification of neural crest lineage and glucose metabolism genes. Importantly, we show acetaldehyde is a metabolite source for acetyl-histone H3 modification in an ALDH1A3, dependent manner providing a physiologic function for this highly volatile and toxic metabolite. In a zebrafish model of melanoma residual disease, a subpopulation of ALDH1-high cells emerges following BRAF inhibitor treatment and targeting these with an ALDH1 suicide inhibitor, nifuroxazide, delays or prevents BRAF inhibitor drug-resistant relapse. Our work reveals that the ALDH1A3-ACSS2 couple directly coordinates nuclear acetaldehyde-acetyl-CoA metabolism with specific chromatin-based gene regulation and represents a potential therapeutic vulnerability in melanoma. HighlightsO_LIALDH1A3-high melanomas are in a high glucose metabolic flux and neural crest stem cell dual state. C_LIO_LINuclear ALDH1A3 partners with ACSS2 to promote selective acetyl-histone H3. C_LIO_LIAcetaldehyde is an acetyl source for ALDH1A3 dependent histone H3 acetylation. C_LIO_LIALDH1A3 is a master regulator and pharmaceutical target for melanoma heterogeneity. C_LI

cancer biology↗

bioRxiv2023

Proteome-scale prediction of molecular mechanisms underlying dominant genetic diseases

Many dominant genetic disorders result from protein-altering mutations, acting primarily through dominant-negative (DN), gain-of-function (GOF), and loss-of-function (LOF) mechanisms. Deciphering the mechanisms by which dominant diseases exert their effects is often experimentally challenging and resource intensive, but is essential for developing appropriate therapeutic approaches. Diseases that arise via a LOF mechanism are more amenable to be treated by conventional gene therapy, whereas DN and GOF mechanisms may require gene editing or targeting by small molecules. Moreover, pathogenic missense mutations that act via DN and GOF mechanisms are more difficult to identify than those that act via LOF using nearly all currently available variant effect predictors. Here, we introduce a tripartite statistical model made up of support vector machine binary classifiers trained to predict whether human protein-coding genes are likely to be associated with DN, GOF, or LOF molecular disease mechanisms. We test the utility of the predictions by examining biologically and clinically meaningful properties known to be associated with the mechanisms. Our results strongly support that the models are able to generalise on unseen data and offer insight into the functional attributes of proteins associated with different mechanisms. We hope that our predictions will serve as a springboard for researchers studying novel variants and those of uncertain clinical significance, guiding variant interpretation strategies and experimental characterisation. Predictions for the human UniProt reference proteome are available at https://osf.io/z4dcp/.

genetics↗

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↗

bioRxiv2023

Correspondence between functional scores from deep mutational scans and predicted effects on protein stability

Many methodologically diverse computational methods have been applied to the growing challenge of predicting and interpreting the effects of protein variants. As many pathogenic mutations have a perturbing effect on protein stability or intermolecular interactions, one highly interpretable approach is to use protein structural information to model the physical impacts of variants and predict their likely effects on protein stability and interactions. Previous efforts have assessed the accuracy of stability predictors in reproducing thermodynamically accurate values and evaluated their ability to distinguish between known pathogenic and benign mutations. Here, we take an alternate approach, and explore how well stability predictor scores correlate with functional impacts derived from deep mutational scanning (DMS) experiments. In this work, we compare the predictions of 9 protein stability-based tools against mutant protein fitness values from 45 independent DMS datasets, covering 161,441 unique single amino acid variants. We find that FoldX and Rosetta show the strongest correlations with DMS-based functional scores, similar to their previous top performance in distinguishing between pathogenic and benign variants. For both methods, performance is considerably improved when considering intermolecular interactions from protein complex structures, when available. Finally, we also highlight that predicted stability effects show consistently higher correlations with certain DMS experimental phenotypes, particularly those based upon protein abundance, and, in certain cases, can be competitive with other sequence-based variant effect prediction methodologies for predicting functional scores from DMS experiments.

bioinformatics↗

bioRxiv2022

Updated benchmarking of variant effect predictors using deep mutational scanning

Variant effect predictors (VEPs) provide a potential solution to the influx of variants of uncertain clinical significance produced by genome sequencing studies. However, the assessment of VEP performance is fraught with biases introduced by benchmarking against clinical observations. In this study, building on our previous work, we use independently generated measurements of protein function from deep mutational scanning (DMS) experiments for 26 human proteins to benchmark 55 different VEPs, while introducing minimum data circularity. The top VEPs are dominated by unsupervised methods including EVE, DeepSequence and ESM-1v, a new protein language model that ranked first overall. However, the strong performance of recent supervised VEPs, in particular VARITY, shows that developers are taking data circularity and bias issues seriously. We also assess the performance of DMS and unsupervised VEPs for discriminating between known pathogenic and putatively benign missense variants. Our findings are mixed, demonstrating that some DMS datasets perform exceptionally at variant classification, while others are poor. Notably, we observe a striking correlation between VEP agreement with DMS data and performance in identifying clinically relevant variants, with EVE, DeepSequence and ESM-1v performing best, further supporting the utility of DMS as an independent benchmark.

bioinformatics↗

bioRxiv2022

Buffering of genetic dominance by allele-specific protein complex assembly

Protein complex assembly often begins while at least one of the subunits is still in the process of being translated. When such cotranslational assembly occurs for homomeric complexes, made up of multiple copies of the same subunit, this will result in complexes whose subunits were translated off of the same mRNA in an allele-specific manner. It has therefore been hypothesised that cotranslational assembly may be able to counter the assembly-mediated dominant-negative effect, whereby the co-assembly of mutant and wild-type subunits "poison" the activity of a protein complex. Here, we address this, showing first that subunits that undergo cotranslational assembly are much less likely to be associated with autosomal dominant relative to recessive disorders. Moreover, we observe that subunits with dominant-negative disease mutations are significantly depleted in cotranslational assembly compared to those associated with loss-of-function mutations. Consistent with this, we also find that complexes with known dominant-negative effects tend to expose their interfaces late during translation, lessening the likelihood of cotranslational assembly. Finally, by combining protein complex properties with other protein-level features, we trained a computational model for predicting proteins likely to be associated with dominant-negative or gain-of-function molecular mechanisms, which we believe will be of considerable utility for protein variant interpretation.

genetics↗

bioRxiv2022

Cell overgrowth during G1 arrest triggers an osmotic stress response and chronic p38 activation to promote cell cycle exit

Cell size and the cell cycle are intrinsically coupled and abnormal increases in cell size are associated with senescence. The mechanism by which overgrowth primes cells to exit the cell cycle remains unclear. We investigate this using CDK4/6 inhibitors that arrest cell cycle progression in G0/G1 and are used to treat ER+/HER2-metastatic breast cancer. We demonstrate that long-term CDK4/6 inhibition promotes cellular overgrowth during the G0/G1 arrest, causing widespread proteome remodeling and p38-p53-p21-dependent cell cycle exit. Cell cycle exit is triggered by two waves of p21 induction. First, overgrowth during a G0/G1 arrest induces an osmotic stress response, producing the first wave of p21 induction. Second, when CDK4/6 inhibitors are removed, a fraction of cells escape G0/G1 arrest and enter S-phase where overgrowth-driven replication stress results in a second wave of p21 induction that causes cell cycle withdrawal from G2, or the subsequent G1. This could explain why cellular hypertrophy is associated with senescence and why CDK4/6 inhibitors have long-lasting anti-proliferative effects in patients.

cell biology↗

bioRxiv2022

Understanding molecular mechanisms and predicting phenotypic effects of pathogenic tubulin mutations

Cells rely heavily on microtubules for several processes, including cell division and molecular trafficking. Mutations in the different tubulin- and -{beta} proteins that comprise microtubules have been associated with various diseases and are often dominant, sporadic and congenital. While the earliest reported tubulin mutations affect neurodevelopment, mutations are also associated with other disorders such as bleeding disorders and infertility. We performed a systematic survey of tubulin mutations across all isotypes in order to improve our understanding of how they cause disease, and increase our ability to predict their phenotypic effects. Both protein structural analyses and computational variant effect predictors were very limited in their utility for differentiating between pathogenic and benign mutations. This was even worse for those genes associated with non-neurodevelopmental disorders. We selected tubulin- and -{beta} disease mutations that were most poorly predicted for experimental characterisation. These mutants co-localise to the mitotic spindle in HeLa cells, suggesting they may exert dominant-negative effects by altering microtubule properties. Our results show that tubulin mutations represent a blind spot for current computational approaches, being much more poorly predicted than mutations in most human disease genes. We suggest that this is likely due to their strong association with dominant-negative and gain-of-function mechanisms.

genetics↗