Research Enablers
QMUL's High-Performance Computing Cluster, Apocrita, enables research across all academic disciplines in the university. Below are a few examples of recent research outputs that have utilised the facility, demonstrating the wide variety of work undertaken. The facility is supported by the ITS Research group at QMUL.
Hybrid Losses for Hierarchical Embedding Learning
"In traditional supervised learning, the cross-entropy loss treats all incorrect predictions equally, ignoring the relevance or proximity of wrong labels to the correct answer. By leveraging a tree hierarchy for fine-grained labels, we investigate hybrid losses, such as generalised triplet and cross-entropy losses, to enforce similarity between labels within a multi-task learning framework. We propose metrics to evaluate the embedding space structure and assess the model’s ability to generalise to unseen classes, that is, to infer similar classes for data belonging to unseen categories. Our experiments on OrchideaSOL, a fourlevel hierarchical instrument sound dataset with nearly 200 detailed categories, demonstrate that the proposed hybrid losses outperform previous works in classification, retrieval, embedding space structure, and generalisation." Read More
Markets versus Morals at Sea: Opportunity Costs and Rescue by Merchant Ships
"This paper examines the interaction between market dynamics and the provision of humanitarian assistance, focusing on merchant-vessel responses to SOS incidents during the Mediterranean migration crisis. Drawing on 1.1 billion ship-positions, we find that commercial vessels do respond; however, they are 90 percent less likely than equally distant NGO-operated vessels to reach those
in distress within the critical first five hours. The opportunity cost of intervening is a key determinant of this muted response." Read More
Exploring the Structural Diversity and Evolution of the D1 Subunit of Photosystem II Using AlphaFold and Foldtree
"Although our knowledge of photosystem II has expanded to include time-resolved atomic details, the diversity of experimental structures of the enzyme remains limited. Recent advances in protein structure prediction with AlphaFold offer a promising approach to fill this gap in structural diversity in non-model systems. This study used AlphaFold to predict the structures of the D1 protein, the core subunit of photosystem II, across a broad range of photosynthetic organisms. The prediction produced high-confidence structures, and structural alignment analyses highlighted conserved regions across the different D1 groups, which were in line with high pLDDT scoring regions. In contrast, varying pLDDT in the DE loop and terminal regions appears to correlate with different degrees of structural flexibility or disorder. " Read More
Contrasting distributions and expression characteristics of transcribing repeats in Setaria viridis
"Repetitive DNA contributes significantly to plant genome size, adaptation, and evolution. However, little is understood about the transcription of repeats. This is addressed here in the plant green foxtail millet (Setaria viridis). First, we used RepeatExplorer2 to calculate the genome proportion (GP) of all repeat types and compared the GP of long terminal repeat (LTR) retroelements against annotated complete and incomplete LTR retroelements (Ty1/copia and Ty3/gypsy) identified by DANTE in a whole genome assembly. We show that DANTE-identified LTR retroelements can comprise ∼0.75% of the inflorescence poly-A transcriptome and ∼0.24% of the stem ribo-depleted transcriptome. In the RNA libraries from inflorescence tissue, both LTR retroelements and DNA transposons identified by RepeatExplorer2 were highly abundant, where they may be taking advantage of the reduced epigenetic silencing in the germ line to amplify." Read More
Exploring Fairness in Long-Term Prediction of Type 2 Diabetes Microvascular Complications
"Existing inequalities are known through out diabetes care which result in poorer health outcomes for ethnic minority groups and those from disadvantaged backgrounds. With the growth of foundation models being deployed to assist with
diagnosis and healthcare usage predictions it is essential we understand how these may exacerbate existing biases."
Read More
Regional and Aggregate Economic Consequences of Environmental Policy
"This paper shows how to combine microeconometric evidence on the effects of environmental policy with a macroeconomic model, accounting for general equilibrium spillovers that have mostly been ignored in the literature. To this end, we study the effects of a recent US air pollution policy. " Read More
Neuroanatomical and prognostic associations of depression in Parkinson’s disease
"Depression is reported as a risk factor, prodromal feature and late consequence of Parkinson's disease (PD). We aimed to evaluate the timing, neuroanatomy and prognostic implications of depression in PD." Read More
Prediction of atrial fibrillation and stroke using machine learning models in UK Biobank
"Atrial fibrillation (AF) is the most common cardiac arrythmia, and it is associated with increased risk for ischemic stroke, which is underestimated, as AF can be asymptomatic. The aim of this study was to develop optimal ML models for prediction of AF in the population, and secondly for ischemic stroke in AF patients." Read More
10 July 2023
Disagreement attention: Let us agree to disagree on computed tomography segmentation
“During the last decade, Deep Learning (DL) has achieved remarkable results on medical datasets when performing tasks like classification, segmentation, object detection, among others. The research on semantic segmentation has led to several Neural Networks (NNs) specialising in creating high-quality masks. The UNet [1] and Deeplab [2] are the most popular. Both are based on Convolutional Neural Networks (CNN) and have been intensively studied and developed for the last seven years.” Read More
Adsorption, activation, and conversion of carbon dioxide on small copper–tin nanoclusters†
"Carbon dioxide (CO2) conversion to value-added chemicals is an attractive solution to reduce globally accelerating CO2 emissions. Among the non-precious and abundant metals tested so far, copper (Cu) is one of the best electrocatalysts to convert CO2 into more than thirty different hydrocarbons and alcohols. However, the selectivity for desired products is often too low. We present a computational investigation of the effects of nanostructuring, doping, and support on the activity and selectivity of Cu–Sn catalysts." Read more
Numerical Investigation of Jet Installation Noise for Chevron Nozzles
"Jet mixing and installation noise remains one of the leading sources of community noise during the take-off phase. Previously, chevron nozzles were primarily used as a noise reduction technology for isolated jet noise. However, a systematic investigation of chevron nozzles in installed configurations is lacking". Read More
Radial-velocity discovery of a second planet in the TOI-1338/BEBOP-1 circumbinary system
"We report the detection of a gas-giant planet in orbit around both stars of an eclipsing binary star system that also contains the smaller, inner transiting planet TOI-1338b. The new planet, called TOI-1338/BEBOP-1c, was discovered using radial-velocity data collected with the HARPS and ESPRESSO spectrographs." Read more
Rigid-Body Sound Synthesis with Differentiable Modal Resonators
"Physical models of rigid bodies are used for sound synthesis in applications from virtual environments to music production. Traditional methods, such as modal synthesis, often rely on computationally expensive numerical solvers, while recent deep learning approaches are limited by post-processing of their results. In this work, we present a novel end-to-end framework for training a deep neural network to generate modal resonators for a given 2D shape and material using a bank of differentiable IIR filters." Read More
Deep palmar phenotyping in atopic eczema: patterns associated with filaggrin variants, disease severity and barrier function in a South Asian population
"Hyperlinear palms are described as a feature of loss-of-function (LoF) variants in filaggrin (FLG). Objectives To explore the phenotype of participants (age < 31 years) with atopic eczema of Bangladeshi ancestry from East London and investigate which factors best associate with LoF FLG variants." Read More
Exploring the “N-Terminal Anchor” Binding Interface of the T3SS Chaperone–Translocator Complexes from P. aeruginosa
"The type III secretion system is a large multiprotein complex that many Gram-negative bacteria use for infection. A crucial part of the complex is its translocon pore formed by two proteins: the major and minor translocators. The pore completes a proteinaceous channel from the bacterial cytosol through the host cell membrane and allows the direct injection of bacterial toxins. Effective pore formation is predicated by the translocator proteins binding to a small chaperone within the bacterial cytoplasm." Read More
11 January 2023
Scientists discover the evolutionary secret behind different animal life cycles
For over 100 years, biologists have wondered why animals display different types of life cycles. Some species, like us humans and most vertebrates, develop directly into a fully formed—yet smaller—version of an adult. In contrast, many other animals give rise to beautifully diverse intermediate forms we call larvae, which then metamorphose into the adult.
In a study led by a team at Queen Mary University of London (QMUL), published in the prestigious journal Nature, researchers uncover for the first time the mechanism that likely explains how embryos form either a larva or a miniature version of the adult.
Nature paper 'Annelid functional genomics reveal the origins of bilaterian life cycles'
10 January 2022
Individual-based modeling of genome evolution in haplodiploid organisms
"Ants, bees, wasps, bark beetles, and other species have haploid males and diploid females. Although such haplodiploid species play key ecological roles and are threatened by environmental changes, no general framework exists for simulating their genetic evolution. Here, we use the SLiM simulation environment to build a novel model for individual-based forward simulation of genetic evolution in haplodiploid populations. We compare the fates of adaptive and deleterious mutations and find that selection is more effective in haplodiploid species than in diploid species. Our open-source model will help understand the evolution of sociality and how ecologically important species may adapt to changing environments." Read More.
Towards a mechanistic understanding of particle shrinkage during biomass pyrolysis via synchrotron X-ray microtomography and in-situ radiography
"Accurate modelling of particle shrinkage during biomass pyrolysis is key to the production of biochars with specific morphologies. Such biochars represent sustainable solutions to a variety of adsorption-dependent environmental remediation challenges. Modelling of particle shrinkage during biomass pyrolysis has heretofore been based solely on theory and ex-situ experimental data. Here we present the first in-situ phase-contrast X-ray imaging study of biomass pyrolysis. A novel reactor was developed to enable operando synchrotron radiography of fixed beds of pyrolysing biomass. Almond shell particles experienced more bulk shrinkage and less change in porosity than did walnut shell particles during pyrolysis, despite their similar composition. Alkaline pretreatment was found to reduce this difference in feedstock behaviour. Ex-situ synchrotron X-ray microtomography was performed to study the effects of pyrolysis on pore morphology. Pyrolysis led to a redistribution of pores away from particle surfaces, meaning newly formed surface area may be less accessible to adsorbates." Read More
A model of large-scale thematic structure
"The coherent organisation of thematic material into large-scale structures within a composition is an important concept in both traditional and cognitive theories of music. However, empirical evidence supporting their perception is scarce. Providing a more nuanced approach, this paper introduces a computational model of hypothesised cognitive mechanisms underlying perception of large-scale thematic structure. Repetition detection based on statistical learning forms the model's foundation, hypothesising that predictability arising from repetition creates perceived thematic coherence. Measures are produced that characterise structural properties of a corpus of 623 monophonic compositions. Exploratory analysis reveals the extent to which these measures vary systematically and independently." Read More
Explosive higher-order Kuramoto dynamics on simplicial complexes
"The higher-order interactions of complex systems, such as the brain are captured by their simplicial complex structure and have a significant effect on dynamics. However the existing dynamical models defined on simplicial complexes make the strong assumption that the dynamics resides exclusively on the nodes. Here we formulate the higher-order Kuramoto model which describes the interactions between oscillators placed not only on nodes but also on links, triangles, and so on. We show that higher-order Kuramoto dynamics can lead to explosive synchronization transition by using an adaptive coupling dependent on the solenoidal and the irrotational component of the dynamics." Read More
Exploiting Class Labels to Boost Performance on Embedding-based Text Classification
"Text classification is one of the most frequent tasks for processing textual data, facilitating among others research from large-scale datasets. Embeddings of different kinds have recently become the de facto standard as features used for text classification. These embeddings have the capacity to capture meanings of words inferred from occurrences in large external collections". Read More
Analysis of the atomic structure of CdS magic-size clusters by X-ray absorption spectroscopy†
"Magic-size clusters are ultra-small colloidal semiconductor systems that are intensively studied due to their monodisperse nature and sharp UV-vis absorption peak compared with regular quantum dots. However, the small size of such clusters (<2 nm), and the large surface-to-bulk ratio significantly limit characterisation techniques that can be utilised. Here we demonstrate how a combination of EXAFS and XANES analyses can be used to obtain information about sample stoichiometry and cluster symmetry." Read More
Endogenous retroviruses are a source of enhancers with oncogenic potential in acute myeloid leukaemia
"Acute myeloid leukemia (AML) is characterised by a series of genetic and epigenetic alterations that result in deregulation of transcriptional networks. One understudied source of transcriptional regulators are transposable elements (TEs), whose aberrant usage could contribute to oncogenic transcriptional circuits. However, the regulatory influence of TEs and their links to AML pathogenesis remain unexplored." Read More
11 August 2020
Rolling Horizon NEAT for General Video Game Playing
"This paper presents a new Statistical Forward Planning (SFP) method, Rolling Horizon NeuroEvolution of Augmenting Topologies (rhNEAT). Unlike traditional Rolling Horizon Evolution, where an evolutionary algorithm is in charge of evolving a sequence of actions, rhNEAT evolves weights and connections of a neural network in real-time, planning several steps ahead before returning an action to execute in the game. Different versions of the algorithm are explored in a collection of 20 GVGAI games, and compared with other SFP methods and state of the art results. Although results are overall not better than other SFP methods, the nature of rhNEAT to adapt to changing game features has allowed to establish new state of the art records in games that other methods have traditionally struggled with. The algorithm proposed here is general and introduces a new way of representing information within rolling horizon evolution techniques. " Read More
6 August 2020
Simplicial complexes: higher-order spectral dimension and dynamics
"Simplicial complexes constitute the underlying topology of interacting complex systems including among the others brain and social interaction networks. They are generalized network structures that allow to go beyond the framework of pairwise interactions and to capture the many-body interactions between two or more nodes strongly affecting dynamical processes. In fact, the simplicial complexes topology allows to assign a dynamical variable not only to the nodes of the interacting complex systems but also to links, triangles, and so on. " Read More
.jpg)
6 August 2020
EVOLUTIONARY COMPUTATION IN COMBINATORIAL OPTIMIZATION: 20th European
"This book constitutes the refereed proceedings of the 20th European Conference on Evolutionary Computation in Combinatorial Optimization, EvoCOP 2020, held as part of Evo*2020, in Seville, Spain, in April 2020, co-located with the Evo*2020 events EuroGP, EvoMUSART and EvoApplications. The 14 full papers presented in this book were carefully reviewed and selected from 37 submissions. The papers cover a wide spectrum of topics, ranging from the foundations of evolutionary computation algorithms and other search heuristics, to their accurate design and application to combinatorial optimization problems." Read More
Peter Pan Discs: finding Neverland’s parameters
"Peter Pan discs are a recently discovered class of long-lived discs around low-mass stars that survive for an order of magnitude longer than typical discs. In this paper we use disc evolutionary models to determine the required balance between initial conditions and the magnitude of dispersal processes for Peter Pan discs to be primordial." Read More
21 November 2019
Genomic basis of European ash tree resistance to ash dieback fungus
"Scientists say there is new hope in the fight against a disease that is devastating ash trees. A study has identified the genes that give trees resistance to ash dieback, which arrived in the UK in 2012 and has now spread to almost every part of the country. The discovery suggests that trees could now be bred that are unaffected by the epidemic. The research is published in the journal Nature Ecology and Evolution." Read More on BBC News
19 November 2019
Cardiovascular Predictive Value and Genetic Basis of Ventricular Repolarization Dynamics
"Early prediction of cardiovascular risk in the general population remains an important issue. The T-wave morphology restitution (TMR), an ECG marker quantifying ventricular repolarization dynamics, is strongly associated with cardiovascular mortality in patients with heart failure. Our aim was to evaluate the cardiovascular prognostic value of TMR in a UK middle-aged population and identify any genetic contribution." Read More
Understanding Aggregation Induced Emission in a Propeller‐Shaped Blue Emitter.
"Organic fluorophores with an enhanced emission in the condensed phase have great potential for the design of optoelectronic materials. Several propeller‐shaped molecules show aggregation‐induced emission (AIE), in particular, silole derivatives have attracted wide attention because of their significant quantum yields in the solid state." Read more
The full dynamics of energy relaxation in large organic molecules: from photo-excitation to solvent heating.
"In some molecular systems, such as nucleobases, polyenes or the active ingredients of sunscreens, substantial amounts of photo-excitation energy are dissipated on a sub-picosecond time scale, raising questions such as: where does this energy go or among which degrees of freedom it is being distributed at such early times? Here we use transient absorption spectroscopy to track excitation energy dispersing from the optically accessible vibronic subsystem into the remaining vibrational subsystem of the solute and solvent. Monitoring the flow of energy during vibrational redistribution enables quantification of local molecular heating". Read more
2 January 2019
Tax evasion as contingent debt.
"This paper studies income-tax evasion in a quantitative incomplete-markets setting with heterogeneous agents. A central aspect is that, realistically, evaded taxes are a form of contingent debt. Since evasion becomes part of a portfolio decision, risk and credit considerations play a central part in shaping it. The model calibrated to match estimated average levels of evasion does a good job in producing observed cross-sectional average evasion rates that decline with age and with earnings.” Read more
Deep learning the high variability and randomness inside multimode fibres.
"This research utilised Queen Mary's Apocrita HPC facility to train our network’s models. The training phase of the network was performed under the Linux Singularity container using a Tesla K80 GPU card (Nvidia) and a 16 Core Xeon E5-2683V3 processor (Intel) with 7.5GB RAM requested for each core.” Read more
18 October 2018
Phenotypically distinct female casts in honey bees are defined by alternative chromatin states during larval development.
“The capacity of the honey bee to produce three phenotypically distinct organisms (two female castes; queens and sterile workers, and haploid male drones) from one genotype represents one of the most remarkable examples of developmental plasticity in any phylum. the queen–worker morphological and reproductive divide is environmentally controlled during post-embryonic development by differential feeding.” Read more
16 February 2018
Ageing-associated DNA methylation dynamics are a molecular readout of lifespan variation among mammalian species.
“Mammalian species exhibit a wide range of lifespans. To date, a robust and dynamic molecular readout of these lifespan differences has not yet been identified. Recent studies have established the existence of ageing-associated differentially methylated positions (aDMPs) in human and mouse.” Read More