Biology subjects

Rooney, M.

Publications and source records attributed to Rooney, M..

7 recordsLinked to original sources

bioRxiv2023

LightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks

The peptide-major histocompatibility complex (pMHC) is a crucial protein in cell-mediated immune recognition and response. Accurate structure prediction is potentially beneficial for protein interaction prediction and therefore helps immunotherapy design. However, predicting these structures is challenging due to the sequential and structural variability. In addition, existing pre-trained models such as AlphaFold 2 require expensive computation thus inhibiting high throughput in silico peptide screening. In this study, we propose LightMHC: a lightweight model (2.2M parameters) equipped with attention mechanisms, graph neural networks, and convolutional neural networks. LightMHC predicts full-atom pMHC structures from amino-acid sequences alone, without template structures. The model achieved comparable or superior performance to AlphaFold 2 and ESMFold (93M and 15B parameters respectively), with five-fold acceleration (6.65 seconds/sample for LightMHC versus 36.82 seconds/sample for AlphaFold 2), potentially offering a valuable tool for immune protein structure prediction and immunotherapy design.

bioinformatics↗

bioRxiv2023

FrameDiPT: SE(3) Diffusion Model for Protein Structure Inpainting

Protein structure prediction field has been revolutionised by deep learning with protein folding models such as AlphaFold 2 and ESMFold. These models enable rapid in silico prediction and have been integrated into de novo protein design and protein-protein interaction (PPI) prediction. However, biologically relevant features dependent on conformational distributions cannot be estimated with these models. Diffusion models, a novel class of generative models, have been developed to learn conformational distributions and applied to de novo protein design. Limited work has been done on protein structure inpainting, where a masked section is recovered by simultaneously conditioning on its sequence and the rest of the structure. In this work, we propose FrameDiff inPainTing (FrameDiPT), a generalised model for protein inpainting. This is important for T-cells given the hyper-variability of the complementarity determining region (CDR) loops. We evaluated the model on CDR loop design for T-cell receptors and achieved comparable prediction accuracy to ProteinGenerator and RFdiffusion with limited training data and learnable parameters. Different from deterministic structure prediction models, FrameDiPT captures the conformational distribution at different regions and binding states, highlighting a key advantage of generative models. The model and inference code have been released1.

immunology↗

bioRxiv2022

Peptide-MHC Structure Prediction With Mixed Residue and Atom Graph Neural Network

Neoantigen-targeting vaccines have achieved breakthrough success in cancer immunotherapy by eliciting immune responses against neoantigens, which are proteins uniquely produced by cancer cells. During the immune response, the interactions between peptides and major histocompatibility complexes (MHC) play an important role as peptides must be bound and presented by MHC to be recognised by the immune system. However, only limited experimentally determined peptide-MHC (pMHC) structures are available, and in-silico structure modelling is therefore used for studying their interactions. Current approaches mainly use Monte Carlo sampling and energy minimisation, and are often computationally expensive. On the other hand, the advent of large high-quality proteomic data sets has led to an unprecedented opportunity for deep learning-based methods with pMHC structure prediction becoming feasible with these trained protein folding models. In this work, we present a graph neural network-based model for pMHC structure prediction, which takes an amino acid-level pMHC graph and an atomic-level peptide graph as inputs and predicts the peptide backbone conformation. With a novel weighted reconstruction loss, the trained model achieved a similar accuracy to AlphaFold 2, requiring only 1.7M learnable parameters compared to 93M, representing a more than 98% reduction in the number of required parameters.

bioinformatics↗

bioRxiv2019

Lipopolysaccharide Simulations are Sensitive to Phosphate Charge and Ion Parameterization

The high proportion of lipopolysaccharide (LPS) molecules in the outer membrane of Gram-negative bacteria make it a highly effective barrier to small molecules, antibiotic drugs, and other antimicrobial agents. Given this vital role in protecting bacteria from potentially hostile environments, simulations of LPS bilayers and outer membrane systems represent a critical tool for understanding the mechanisms of bacterial resistance and the development of new antibiotic compounds that circumvent these defenses. The basis of these simulations are parameterizations of LPS, which have been developed for all major molecular dynamics force fields. However, these parameterizations differ in both the protonation state of LPS as well as how LPS membranes behave in the presence of various ion species. To address these discrepancies and understand the effects of phosphate charge on bilayer properties, simulations were performed for multiple distinct LPS chemotypes with different ion parameterizations in both protonated or deprotonated lipid A states. These simulations show that bilayer properties, such as the area per lipid and inter-lipid hydrogen bonding, are highly influenced by the choice of phosphate group charges, cation type, and ion parameterization, with protonated LPS and monovalent cations with modified nonbonded parameters providing the best match to experiments. Additionally, alchemical free energy simulations were performed to determine theoretical pKa values for LPS, and subsequently validated by 31P solid-state NMR experiments. Results from these complementary computational and experimental studies demonstrate that the protonated state dominates at physiological pH, contrary to the deprotonated form modeled by many LPS force fields. In all, these results highlight the sensitivity of LPS simulations to phosphate charge and ion parameters, while offering recommendations for how existing models should be updated for consistency between force fields as well as to best match experiments.

biophysics↗

bioRxiv2019

Circulating electrolytes and the prevalence of atrial fibrillation and supraventricular ectopy: the Atherosclerosis Risk in Communities (ARIC) Study

BackgroundEvaluating associations of circulating electrolytes with atrial fibrillation (AF) and burden of supraventricular arrhythmias can give insights into arrhythmia pathogenesis.\n\nMethodsWe conducted a cross-sectional analysis of 6,398 participants of the Atherosclerosis Risk in Communities (ARIC) study, ages 71-90, with data on serum electrolytes (magnesium, calcium, potassium, phosphorus, chloride, sodium). Prevalence of AF was determined from study 12-lead electrocardiograms and prior history of AF-related hospitalizations. A subset of 317 participants also underwent electrocardiographic recordings for up to 14 days using the Zio(R) patch. Burden of other supraventricular arrhythmias [premature atrial contractions (PACs), supraventricular tachycardia] was determined with the Zio(R) patch. We used multivariable logistic and linear regression adjusting for potential confounders to determine associations of electrolytes with arrhythmia prevalence and burden.\n\nResultsAmong 6,394 eligible participants, 614 (10%) had prevalent AF. Participants in the top quintiles of magnesium [odds ratio (OR) 0.82, 95% confidence interval (CI) 0.62, 1.08], potassium (OR 0.82, 95%CI 0.68, 1.00), and phosphorus (OR 0.73, 95%CI 0.59, 0.89) had lower prevalence of AF compared to those in the bottom quintiles. No clear association was found for circulating chloride, calcium or sodium. Higher concentrations of circulating calcium were associated with lower prevalence of PACs using a standard 12-lead electrocardiogram, while higher concentrations of potassium, chloride and sodium were associated with higher PAC prevalence. Circulating electrolytes were not significantly associated with the burden of PACs or supraventricular tachycardia among a subset of 317 participants with extended electrocardiographic monitoring.\n\nConclusionConcentrations of circulating electrolytes present complex associations with selected supraventricular arrhythmias. Future studies should evaluate underlying mechanisms.

epidemiology↗

bioRxiv2018

Comprehensive analysis of tumour initiation, spatial and temporal progression under multiple lines of treatment

Driver mutations alter cells from normal to cancer through several evolutionary epochs: premalignancy, early malignancy, subclonal diversification, metastasis and resistance to therapy. Later stages of disease can be explored through analyzing multiple samples collected longitudinally, on or between successive treatments, and finally at time of autopsy. It is also possible to study earlier stages of cancer development through probabilistic reconstruction of developmental trajectories based on mutational information preserved in the genome. Here we present a suite of tools, called Phylogic N-Dimensional with Timing (PhylogicNDT), that statistically model phylogenetic and evolutionary trajectories based on mutation and copy-number data representing samples taken at single or multiple time points. PhylogicNDT can be used to infer: (i) the order of clonal driver events (including in pre-cancerous stages); (ii) subclonal populations of cells and their phylogenetic relationships; and (iii) cell population dynamics. We demonstrate the use of PhylogicNDT by applying it to whole-exome and whole-genome data of 498 lung adenocarcinoma samples (434 previously available and 64 of newly generated data). We identify significantly different progression trajectories across subtypes of lung adenocarcinoma (EGFR mutant, KRAS mutant, fusion-driven and EGFR/KRAS wild type cancers). In addition, we study the progression of fusion-driven lung cancer in 21 patients by analyzing samples from multiple timepoints during treatment with 1st and next generation tyrosine kinase inhibitors. We characterize their subclonal diversification, dynamics, selection, and changes in mutational signatures and neoantigen load. This methodology will enable a systematic study of tumour initiation, progression and resistance across cancer types and therapies.

bioinformatics↗

bioRxiv2018

Microglial and Astrocyte priming in the APP/PS1 model of Alzheimer’s Disease: increased vulnerability to acute inflammation and cognitive deficits

Alzheimers disease (AD) causes devastating cognitive decline and has no disease-modifying therapies. Neuroinflammation is a significant contributor to disease progression but its precise contribution remains unclear. An emerging literature indicates that secondary inflammatory insults including acute trauma and infection alter the trajectory of chronic neurodegenerative diseases and the roles of microglia and astrocytes require elucidation. The current study, using the APP/PS1 mouse model of AD, demonstrates that microglia are primed by {beta}-amyloid pathology to induce exaggerated IL-1{beta} responses to acute stimulation with LPS or IL-1{beta}. Despite disease-associated NLRP3 inflammasome activation, evidenced by ASC speck formation, APP/PS1 microglial cells show neither IL-1{beta} induction nor NF{kappa}B p65 nuclear localisation. Upon secondary stimulation with LPS or IL-1{beta}, NF{kappa}B-p65 nuclear localisation and exaggerated pro-IL-1 induction occur. Microglial priming was also unmasked by secondary stimulation with systemic LPS leading to significant cognitive impairment in APP/PS1 mice compared to WT LPS-treated mice. Astrocytes have also recently emerged as displaying significant phenotypic heterogeneity. Here, by-passing microglial priming, and acutely challenging mice with intra-hippocampal IL-1{beta} we demonstrate that astrocytes proximal to A{beta}-plaques are also primed to produce exaggerated CCL2, CXCL1 and CXCL10 responses. Many astrocytosis-associated genes in APP/PS1 mice share these exaggerated responses to IL-1{beta}, while others are equally induced in both strains. Collectively the data show that the amyloid-laden brain shows multiple vulnerabilities to secondary inflammatory challenge: both microglia and astrocytes are primed to produce exaggerated secondary inflammation and systemic LPS is sufficient to cause cognitive impairments relevant to delirium, selectively in animals with prior amyloid pathology.

neuroscience↗