E02-JEM ARM 300CF
|
Diamond Proposal Number(s):
[33252]
Open Access
Abstract: The Noise2Void technique is demonstrated for successful denoising of atomic resolution scanning transmission electron microscopy (STEM) images. The technique is applied to denoising atomic resolution images and videos of gold adatoms on a graphene surface within a graphene liquid-cell, with the denoised experimental data qualitatively demonstrating improved visibility of both the Au adatoms and the graphene lattice. The denoising performance is quantified by comparison to similar simulated data and the approach is found to significantly outperform both total variation and simple Gaussian blurring. Compared to other denoising methods, the Noise2Void technique has the combined advantages that it requires no manual intervention during training or denoising, no prior knowledge of the sample and is compatible with real-time data acquisition rates of at least 45 frames per second.
|
Jan 2026
|
|
|
|
Open Access
Abstract: The optimisation of cooling system performance in tokamak reactors requires accurate analysis of thermo-hydraulic interactions in complex flow regimes. Although Computational Fluid Dynamics (CFD) provides high-fidelity insights at lower cost than experimental testing, optimisation remains computationally prohibitive due to the large number of simulations required. To address this, we introduce Hammerhead, an open-source framework that integrates CFD with surrogate-based optimisation for pipe flow heat and mass transfer problems. Hammerhead constructs a parametric database of high-fidelity simulations by deforming pipe wall geometries using up to four shape parameters. Surrogate models, constructed on the basis of radial basis function interpolation, feed-forward neural networks, and Gaussian processes, are trained on this database to approximate thermo-hydraulic responses efficiently. The framework provides a comparative environment for systematically assessing surrogate model accuracy and reliability within the same optimisation workflow. Implemented in Python 3 and interfaced with OpenFOAM, Hammerhead enables reproducible, high-fidelity CFD studies combined with Machine Learning-aided optimisation. This approach reduces computational cost while maintaining accuracy, offering a practical tool for the design and optimisation of cooling systems in fusion reactor applications.
|
Dec 2025
|
|
I13-2-Diamond Manchester Imaging
|
Kunning
Tang
,
Ryan T.
Armstrong
,
Peyman
Mostaghimi
,
Yufu
Niu
,
Quentin
Meyer
,
Chuan
Zhao
,
Donal P.
Finegan
,
Melissa
Popeil
,
Kamaljit
Singh
,
Hannah
Menke
,
Alexandros
Patsoukis Dimou
,
Tom
Bultreys
,
Arjen
Mascini
,
Mark
Knackstedt
,
Ying
Da Wang
Open Access
Abstract: The recent introduction of deep learning methods for image processing has greatly advanced the characterization of materials using three-dimensional (3D) X-ray imaging techniques. However, deep learning models often have difficulty performing consistently across images owing to unavoidable variations in imaging conditions, which create inconsistencies even for the same material. As a result, networks must frequently be retrained for new datasets, limiting their applicability and generalization. Thus, it is critical to reduce the variations between images to enable a single model to process multiple datasets. Herein, we introduce P3T-Net, a pseudo-3D domain transfer network that transfers diverse 3D images into a uniform domain before processing using deep learning models. Remarkably, P3T-Net enables the reuse of previously trained networks for processing new images and considerably reduces the computational cost of transferring 3D images across domains. These unique capabilities were demonstrated in the following scenarios: (i) image enhancement of fast scans for geological rock and hydrogen fuel cells, (ii) enhancement of images to match the quality of multi-source imaging for lithium-ion batteries, (iii) accurate segmentation of images captured under different conditions, and (iv) tera-scale 3D transfer (1011 voxels) on a single GPU. Overall, the proposed approach addresses cross-domain inconsistencies across various materials and conditions, thereby enabling more robust and generalizable deep learning solutions for a wide range of material imaging tasks.
|
Dec 2025
|
|
I24-Microfocus Macromolecular Crystallography
|
Diamond Proposal Number(s):
[31800]
Open Access
Abstract: Solenoid proteins are elongated tandem repeat proteins with diverse biological functions, making them attractive targets for protein design. Advances in machine learning have transformed our understanding of sequence-structure relationships, enabling new approaches for de novo protein design. Here, we present an in silico evolution platform that couples a solenoid discriminator network with AlphaFold2 as an oracle within a genetic algorithm. Starting from random sequences, we design α-, β-, and αβ-solenoid backbones, generating structures that span natural and novel solenoid space. We experimentally characterise 41 solenoid designs, with α-solenoids consistently folding as intended, including one structurally validated design that closely matches the design model. All β-solenoids initially failed, reflecting the difficulty of designing β-strand majority proteins. By introducing terminal capping elements and refining designs based on earlier experimental screens, we generate two β-solenoids that have biophysical properties consistent with their designs. Our approach achieves fold-specific hallucination-based design without depending on explicit structural templates.
|
Dec 2025
|
|
I12-JEEP: Joint Engineering, Environmental and Processing
|
Abstract: This thesis develops and evaluates a fully automated deep learning workflow for segmenting voids and material phases in X-ray computed tomography of carbon fibre and flax fibre polymer composites. For the carbon fibre system, fibre, matrix, and void masks were generated through a reproducible Python-based pipeline combining slice-wise intensity normalisation, two-stage void detection, morphological refinement, and Otsu-based fibre–matrix separation, followed by targeted manual verification in Dragonfly. For the flax fibre system, segmentation masks were produced and refined manually and unified into a consistent three-phase labelling scheme (matrix, fibre, void). These reference datasets were then used to train five lightweight convolutional neural network architectures (UNet++, UNet3+, Attention UNet, DeepLabV3+ Transformer, and LR-ASPP Transformer) under identical conditions, using paired 256 × 256 patches, controlled label-safe augmentation, and a fixed slice-level train–validation split. A higher-load carbon fibre state (140 N) was withheld entirely from training and used exclusively to evaluate the models on previously unseen microstructural damage. The results demonstrate that compact encoder–decoder networks can accurately localise voids and robustly separate fibre and matrix phases across both composite systems, including under low contrast and evolving damage, while maintaining computational efficiency suitable for routine XCT workflows.
|
Nov 2025
|
|
I03-Macromolecular Crystallography
|
Yazhou
Wang
,
Xiaomin
Wang
,
Tingting
Liu
,
Chao
Wang
,
Qingshuo
Meng
,
Fanye
Meng
,
Jiaojiao
Yu
,
Jinxin
Liu
,
Yaya
Fan
,
David
Gennert
,
Frank W.
Pun
,
Alex
Aliper
,
Feng
Ren
,
Man
Zhang
,
Xin
Cai
,
Xiao
Ding
,
Alex
Zhavoronkov
Open Access
Abstract: PKMYT1 has recently emerged as a compelling therapeutic target for precision cancer therapy due to its synthetic lethality with oncogenic alterations such as CCNE1 amplification and mutations in FBXW7 and PPP2R1A. Current small molecule PKMYT1 inhibitors face limitations, such as insufficient molecular diversity and poor selectivity. We herein use our generative AI platform to develop a bifunctional PKMYT1 degrader by linking an entirely novel PKMYT1 inhibitor to an optimized cereblon (CRBN) binder. The lead PROTAC D16-M1P2 demonstrates dual mechanisms of PKMYT1 degradation and inhibition, with strong antiproliferative potency facilitated by high selectivity. It also exhibits favorable oral bioavailability, stronger pharmacodynamic effects relative to the PKMYT1 inhibitor alone, and robust antitumor response as a monotherapy in xenograft models. This PROTAC serves as a precise chemical probe to explore PKMYT1 biology and a promising lead for further cancer therapy exploration.
|
Nov 2025
|
|
I13-2-Diamond Manchester Imaging
|
Diamond Proposal Number(s):
[22053, 31714]
Abstract: Accurate predictions of the size and morphology of microstructural features, including defects such as porosity, are essential for predicting the performance of engineering components. Although several multiscale approaches exist in the literature, including direct simulations and volume-averaged models, their predictions are limited due to large computational times and relatively low accuracy. This work utilises transfer learning to link the macroscopic field variable distributions to the mesoscale, in order to estimate sub-grid microstructural defects. Specifically, the model parameters are corrected using experimental measurements of sub-grid scale defects. The proposed methodology is illustrated for predicting porosity in an aluminium alloy automotive component produced using high pressure die casting. The model uses a physics-based localised porosity model for combined gas and shrinkage porosity to train an artificial neural network. This trained machine learning model is subsequently re-trained using macroscale field variables and experimental X-ray microtomography porosity measurements from industrial component made using different process conditions. An unseen region of the same component is used for further testing of the performance of the model. The results show good prediction of pore size distribution and location. These results are then used to determine component fatigue life. Thus, a full process-structure-property model is established. The framework has the potential to be applied to a large class of problems involving predictions of microstructural features over entire macroscopic components.
|
Nov 2025
|
|
DIAD-Dual Imaging and Diffraction Beamline
I12-JEEP: Joint Engineering, Environmental and Processing
|
Franck P.
Vidal
,
Shaghayegh
Afshari
,
Sharif
Ahmed
,
Alberto
Albiol
,
Francisco
Albiol
,
Éric
Béchet
,
Alberto Corbí
Bellot
,
Stefan
Bosse
,
Simon
Burkhard
,
Younes
Chahid
,
Cheng-Ying
Chou
,
Robert
Culver
,
Pascal
Desbarats
,
Lewis
Dixon
,
Johan
Friemann
,
Amin
Garbout
,
Marcos
García-Lorenzo
,
Jean-François
Giovannelli
,
Ross
Hanna
,
Clémentine
Hatton
,
Audrey
Henry
,
Graham
Kelly
,
Christophe
Leblanc
,
Alberto
Leonardi
,
Jean Michel
Létang
,
Harry
Lipscomb
,
Tristan
Manchester
,
Bas
Meere
,
Claire
Michelet
,
Simon
Middleburgh
,
Radu P.
Mihail
,
Iwan
Mitchell
,
Liam
Perera
,
Martí
Puig
,
Malek
Racy
,
Ali
Rouwane
,
Hervé
Seznec
,
Aaron
Sújar
,
Jenna
Tugwell-Allsup
,
Pierre-Frédéric
Villard
Diamond Proposal Number(s):
[29820]
Open Access
Abstract: gVirtualXray (gVXR) is an open-source framework that relies on the Beer–Lambert law to simulate X-ray images in real time on a graphics processor unit (GPU) using triangular meshes. A wide range of programming languages is supported (C/C++, Python, R, Ruby, Tcl, C#, Java, and GNU Octave). Simulations generated with gVXR have been benchmarked with clinically realistic phantoms (i.e. complex structures and materials) using Monte Carlo (MC) simulations, real radiographs and real digitally reconstructed radiographs (DRRs), and X-ray computed tomography (XCT). It has been used in a wide range of applications, including real-time medical simulators, proposing a new densitometric radiographic modality in clinical imaging, studying noise removal techniques in fluoroscopy, teaching particle physics and X-ray imaging to undergraduate students in engineering, and XCT to masters students, predicting image quality and artifacts in material science, etc. gVXR has also been used to produce a high number of realistic simulated images in optimisation problems and to train machine learning algorithms. This paper presents a comprehensive review of such applications of gVXR.
|
Nov 2025
|
|
|
|
Kang
Xiang
,
Ling
Qin
,
Shi
Huang
,
Hongyuan
Song
,
Vasilii
Bazhenov
,
Sarlota
Birnšteinová
,
Raphael
De Wijn
,
Jayanath C. P.
Koliyadu
,
Faisal H. M.
Koua
,
Adam
Round
,
Ekaterina
Round
,
Abhisakh
Sarma
,
Tokushi
Sato
,
Marcin
Sikorski
,
Yuhe
Zhang
,
Eleni
Myrto Asimakopoulou
,
Pablo
Villanueva-Perez
,
Kyriakos
Porfyrakis
,
Iakovos
Tzanakis
,
Dmitry G.
Eskin
,
Nicole
Grobert
,
Adrian
Mancuso
,
Richard
Bean
,
Patrik
Vagovic
,
Jiawei
Mi
,
Valerio
Bellucci
Open Access
Abstract: Using megahertz x-ray free electron laser imaging with x-ray pulses of ~25 femtoseconds and a machine-learning strategy, we have conducted comprehensive in situ imaging studies on the dynamics of cavitation bubble clouds in ultrasound fields at the SPB/SFX beamline of the European XFEL. The research unambiguously revealed the quasi-simultaneous implosion of multiple bubbles and simultaneous collapse of bubble cloud in nanosecond scale and their dynamic impacts onto two-dimensional (2D) materials for layer exfoliation. We have also performed multiphysics modeling to simulate the shock wave emission, propagation, impact, and stresses produced. We elucidated the critical conditions for producing instant or fatigue exfoliation and the effects of bonding strengths and structural defects on the exfoliation rate. The discoveries have filled the long-standing missing knowledge gaps in the underlying physics of exfoliating 2D materials in ultrasound fields, providing a solid theoretical foundation for optimizing and scaling-up operation to produce 2D materials in a much more cost-effective and sustainable way.
|
Nov 2025
|
|
|
|
Beatriz
Costa-Gomes
,
Joel
Greer
,
Nikolai
Juraschko
,
James
Parkhurst
,
Jola
Mirecka
,
Marjan
Famili
,
Camila
Rangel-Smith
,
Oliver
Strickson
,
Alan
Lowe
,
Mark
Basham
,
Tom
Burnley
Open Access
Abstract: Ease of access to data, tools and models expedites scientific research. In structural biology there are now numerous open repositories of experimental and simulated data sets. Being able to easily access and utilize these is crucial to allow researchers to make optimal use of their research effort. The tools presented here are useful for collating existing public cryoEM data sets and/or creating new synthetic cryoEM data sets to aid the development of novel data processing and interpretation algorithms. In recent years, structural biology has seen the development of a multitude of machine-learning-based algorithms to aid numerous steps in the processing and reconstruction of experimental data sets and the use of these approaches has become widespread. Developing such techniques in structural biology requires access to large data sets, which can be cumbersome to curate and unwieldy to make use of. In this paper, we present a suite of Python software packages, which we collectively refer to as PERC (profet, EMPIARreader and CAKED). These are designed to reduce the burden which data curation places upon structural biology research. The protein structure fetcher (profet) package allows users to conveniently download and cleave sequences or structures from the Protein Data Bank or AlphaFold databases. EMPIARreader allows lazy loading of Electron Microscopy Public Image Archive data sets in a machine-learning-compatible structure. The Class Aggregator for Key Electron-microscopy Data (CAKED) package is designed to seamlessly facilitate the training of machine-learning models on electron microscopy data, including electron-cryo-microscopy-specific data augmentation and labeling. These packages may be utilized independently or as building blocks in workflows. All are available in open-source repositories and designed to be easily extensible to facilitate more advanced workflows if required.
|
Oct 2025
|
|
