Tag: Joana B. Pereira

Global graph features unveiled by unsupervised deep learning published in Machine Learning: Science and Technology

GAUDI leverages a hierarchical graph-convolutional variational autoencoder architecture, where an encoder progressively compresses the graph into a low-dimensional latent space, and a decoder reconstructs the graph from the latent embedding. (Image by M. Granfors and J. Pineda.)
Global graph features unveiled by unsupervised deep learning
Mirja Granfors, Jesús Pineda, Blanca Zufiria Gerbolés, Joana B. Pereira, Carlo Manzo, Giovanni Volpe
Machine Learning: Science and Technology, 7, 045054 (2026)
arXiv: 2503.05560
doi: 10.1088/2632-2153/ae8d7f

Graphs provide a powerful framework for modeling complex systems, but their structural variability poses significant challenges for analysis and classification. To address these challenges, we introduce Graph Autoencoder Uncovering Descriptive Information (GAUDI), an unsupervised graph deep learning framework designed to capture both local details and global structure. GAUDI employs an hourglass architecture with hierarchical pooling and upsampling layers linked through skip connections, which preserve essential connectivity information throughout the encoding–decoding process. Even though identical or highly similar underlying parameters describing a system’s state can lead to significant variability in graph realizations, GAUDI consistently maps them into nearby regions of a structured and continuous latent space, effectively disentangling invariant process-level features from stochastic noise. We demonstrate GAUDI’s versatility across multiple applications, including small-world networks modeling, characterization of protein assemblies from super-resolution microscopy, analysis of collective motion in the Vicsek model, and identification of age-related changes in brain connectivity. Comparison with related approaches highlights GAUDI’s superior performance in analyzing complex graphs, providing new insights into emergent phenomena across diverse scientific domains.

Vulnerability of locus coeruleus connections to aging and Alzheimer’s disease published in Alzheimer’s & Dementia

Anatomical gradients of the right (R) and left (L) LC in the AD continuum cohort (ADNI). (Image from the article.)
Vulnerability of locus coeruleus connections to aging and Alzheimer’s disease
Blanca Zufiria-Gerbolés, Daniel Vereb, Mite Mijalkov, Massimiliano Passaretti, Giovanni Volpe, Zhilei Xu, Thomas Hinault, Sara Garcia-Ptacek, Joana B. Pereira, for the Alzheimer’s Disease Neuroimaging Initiative
Alzheimer’s & Dementia 22, e71737 (2026)
doi: 10.1002/alz.71737

INTRODUCTION
The locus coeruleus (LC) is important in coordinating communication between brain regions through its widespread connections. However, the organization of its connections, and its changes with aging and Alzheimer’s disease (AD), remains unclear.

METHODS
We mapped whole-brain white matter connections from the LC in two independent cohorts: one spanning the adult lifespan and another covering the AD continuum.

RESULTS
We identified a novel dorsal–ventral organization of LC connectivity that showed changes common to aging and AD or specific to AD, was linked to gene expression patterns, and was associated with cognitive performance and tau pathology in the entorhinal cortex. We also developed an imaging marker, LC gap, capturing LC connectivity deviations associated with slower tau accumulation in cognitively normal individuals at high risk for AD.

DISCUSSION
These findings provide new insights into LC connectivity in aging and AD, highlighting its potential role as a resilience marker against AD neurodegeneration.

Highlights

  • For the first time, we mapped the whole-brain white matter connections from the Locus Coeruleus (LC) to the rest of the brain across aging and Alzheimer’s disease (AD) continuum.
  • Revealed a novel dorsal–ventral organization LC connectivity that presents changes common to aging and AD or specific to AD.
  • LC connectivity patterns are associated with cognitive performance and tau pathology in the entorhinal cortex.
  • We introduced a novel marker capturing LC connectivity deviations and that is associated with tau accumulation in cognitively normal individuals at high risk for AD.

Protein Dynamics Beyond Structure Prediction on ArXiv

Computational advances in protein folding studies. Current approaches address multiple levels of resolution and methodological frameworks, however, none of the existing methods provides quantitative and dynamic information of the relationship between protein sequence and folding mechanism at all-atom resolution and at scale. (Graphics by J. Sacquegno.)
Protein Dynamics Beyond Structure Prediction
Juliette Griffié, Sviatlana Shashkova, Antonio Ciarlo, Sreekanth K. Manikandan, Claes Andréasson, Malin Bäckström, Tristan Bereau, Hjalmar Brismar, Carlos Bustamante, Marta Carroni, Roberto Covino, Andreas Dahlin, Sebastian Deindl, Lucie Delemotte, Arne Elofsson, John Eriksson, Giovanna Fragneto, Anders Gunnarsson, Per Hammarström, Caroline Ingre, Christian Kaiser, Petronella Kettunen, Mark C. Leake, Benjamin Loos, Anna Månberg, Antonia S. J. S. Mey, Richard Neutze, Thomas Nyström, Karl Palmås, Charley Schaefer, Markus J. Tamás, Nicola Ticozzi, Tomás S. Pilvelic, Jacopo Sacquegno, B.M. (Betty)Tijms, Gunnar von Heijne, Björn Wallner, Vitali Zhaunerchyk, Simon Olsson, Joana B. Pereira, Julia Fernandez-Rodriguez, Fredrik Westerlund, Giovanni Volpe
arXiv: 2606.08647

How microorganisms respond to and interact with their environment can vary significantly from individual to individual, which can have important microbiological and ecological implications. However, most microscopy techniques can only observe motile microorganisms for short times because of their limited fields of view. Using Lagrangian tracking, a single microorganism can be followed in 3D, potentially indefinitely, allowing to decipher individual phenotypical traits. Current Lagrangian tracking methods use the fluorescence signal emitted by the microorganism as feedback to keep it in focus. However, over long times, epifluorescent imaging can induce photobleaching and photodamage, and importantly, not all microorganisms can easily be made fluorescent. Additionally, traditional algorithms used in feedback loops to determine microorganism position are prone to errors, especially in optically complex media. Here, we present a faster, more reliable, and versatile Lagrangian tracking method that uses deep learning to determine the 3D position of the microorganism. This new method demonstrates enhanced accuracy and speed in tracking fluorescent bacteria with fluorescence microscopy also in optically complex media. Furthermore, we track bacteria with other microscopy modalities, such as brightfield microscopy — for example, this enables us to track magnetotactic bacteria, which cannot be made fluorescent without degrading their magnetotactic properties. These novel capabilities allow to extract previously inaccessible quantitative information, significantly advancing the study of microorganism behavior — and thus opening new avenues for research in complex biological and ecological systems.

Three-dimensional quantitative tissue clearing reveals differences in osteovascular niche of aged and young human mesenchymal stromal cells published in Nature Biomedical Engineering

Visualization of the vasculature within human bone from a 75-year-old patient by immunostaining with antibodies against CD31. (Image from the manuscript.)
Three-dimensional quantitative tissue clearing reveals differences in osteovascular niche of aged and young human mesenchymal stromal cells
Nelson Tsz Long Chu, Ostap Dregval, Yu-Wei Chang, Emil Kriukov, Xin Tian, Xin Liu, Dana Trompet, Misty Shuo Zhang, Lei Li, Zhong Li, Emiliano Gomez Ruiz, Joana B. Pereira, Mats Brittberg, Björn Barenius, Lars Sävendahl, Ralf H. Adams, Inger Gjertsson, Claes Ohlsson, Giovanni Volpe & Andrei S. Chagin
Nature Biomedical Engineering (2026)
bioRxiv: 10.1101/2025.10.07.680053
doi: 10.1038/s41551-026-01645-3

Human bone marrow mesenchymal stromal/stem cells (BM-MSCs) are widely used in clinical trials and tissue engineering, yet their native microenvironment remains poorly understood. Here we introduce a tissue-clearing protocol, DeepBone, for human bones and integrate it with simultaneous mRNA and protein detection. Using this protocol, we spatially map BM-MSCs relative to key bone microenvironment components, including human blood capillaries, adipocytes, sinusoids and bony trabeculae. Quantitative analysis reveals that the native microenvironment of human BM-MSCs in young bone is enriched in vasculature, sinusoids, bone matrix and adipocytes. In contrast, in aged bone, BM-MSCs show no preferential association with bone or adipocytes. Proliferative BM-MSCs are predominantly found along blood vessels. Moreover, we identify a specialized microenvironment for BM-MSCs in young bone, characterized by sinusoids coiled around trabeculae and enriched by R-type vessels. These findings provide insights into the native niches of BM-MSCs, offering a foundation for the development of tissue engineering strategies that mimic their physiological context.

Mapping individual molecular connectomes in Alzheimer’s disease published in Alzheimer’s & Dementia

Diagnostic classification. (Image from the article.)
Mapping individual molecular connectomes in Alzheimer’s disease
Zhilei Xu, Mite Mijalkov, Jiawei Sun, Yu-Wei Chang, Arianna Sala, Giovanni Volpe, Mario Severino, Mattia Veronese, Sara Garcia-Ptacek, Joana B. Pereira, for the Alzheimer’s Disease Neuroimaging Initiative
Alzheimer’s & Dementia 22, e71310 (2026)
DOI: 10.1002/alz.71310

INTRODUCTION
Mapping individual differences is crucial to improve personalized medicine approaches in Alzheimer’s disease (AD), which is characterized by strong inter-individual variability in the accumulation patterns of tau and amyloid beta pathology.

METHODS
We assess the progression of AD across the disease continuum by building individual molecular connectomes using longitudinal positron emission tomography (PET) data.

RESULTS
We demonstrate that these connectomes constitute a unique fingerprint, capable of identifying a single individual from a large group of subjects. Alterations in the connectomes discriminate different diagnostic groups and predict cognitive decline to a higher extent than conventional PET measures. We introduce a novel gene-specific transcription network analysis that linked individual tau and amyloid connectomes to a common transcriptomic profile of apoptosis, with the tau connectome being specifically related to pyrimidine metabolism, and the amyloid connectome to histone acetylation.

DISCUSSION
Individual molecular connectome mapping provides a novel and sensitive framework to monitor AD progression.

Highlights

  • Individual molecular connectomes constitute a unique fingerprint, capable of identifying a single individual from a large group of subjects.
  • Alterations in individual molecular connectomes significantly increase both across the Alzheimer’s disease (AD) continuum and over time.
  • Alterations in individual molecular connectomes discriminate different diagnostic groups and predict cognitive decline to a higher extent than conventional positron emission tomography measures.
  • Susceptibilities of individual tau and amyloid connectomes to AD are linked to a common transcriptomic profile of apoptosis, with the tau connectome being specifically related to pyrimidine metabolism, and the amyloid connectome to histone acetylation.

Tracking early cognitive decline in preclinical AD with brain MRI similarity published in Alzheimer’s & Dementia

Parcellation of the brain cortex. (Image from the article.)
Tracking early cognitive decline in preclinical AD with brain MRI similarity
Jiawei Sun, Blanca Zufiria-Gerbolés, Massimiliano Passaretti, Giovanni Volpe, Mite Mijalkov, Joana B. Pereira, for the Alzheimer’s Disease Neuroimaging Initiative
Alzheimer’s & Dementia 22, e71170 (2026)
DOI: 10.1002/alz.71170

INTRODUCTION
Early detection of neuroanatomical changes in preclinical Alzheimer’s disease (AD) is critical for timely intervention. However, conventional magnetic resonance imaging (MRI) and fluid biomarkers often lack sensitivity to subtle structural alterations in early disease stages.

METHODS
To identify early brain alterations, we applied a perturbation-based brain similarity approach to cognitively normal participants from Alzheimer’s Disease Neuroimaging Initiative (ADNI) and Open Access Series of Imaging Studies (OASIS), stratified by amyloid status. We evaluated its predictive performance for cognition and diagnostic conversion against cortical thickness, volumetric MRI, and fluid biomarkers.

RESULTS
In both cohorts, brain similarity consistently outperformed other biomarkers across cognitive domains and amyloid groups. It also achieved superior accuracy in predicting clinical conversion and exhibited associations with cytoarchitectural organization.

DISCUSSION
These findings highlight brain similarity as a sensitive marker of early neuroanatomical disruption in AD. Its ability to detect subtle structural changes before overt atrophy underscores its potential for early disease monitoring and treatment assessment in preclinical AD trials.

Highlights

  • Brain similarity captures early brain changes in preclinical Alzheimer’s disease (AD).
  • Brain similarity outperforms conventional biomarkers such as cortical thickness, volume measures, and fluid biomarkers in predicting cognitive decline.
  • Brain similarity predicts conversion to mild cognitive impairment and AD more accurately than traditional imaging markers, and its predictive performance is further improved when combined with fluid biomarkers.
  • Brain similarity captures structural disruptions associated with cortical layer II of the cytoarchitectonic lamina of human neocortex.

International conference “Protein Folding in Real Time: From Molecules to Disease”, Aula Medica, KI, Stockholm, 11-13 March 2026

Giovanni Volpe opens the Protein Folding in Real Time conference. (Photo by A. Ciarlo)
The international conference Protein Folding in Real Time: From Molecules to Disease opened today, 11 March 2026, at Aula Medica, KI, Stockholm.

The conference brings together researchers from multiple disciplines, including biophysics, molecular biology, computational science, and medicine, to discuss recent advances in the study of protein folding. Proteins must fold into precise three-dimensional structures to perform their biological functions, and failures in this process are associated with several diseases, including neurodegenerative disorders and cancer.

During the three-day meeting, participants attend a series of lectures and discussions covering topics such as single-molecule biophysics, high-resolution experimental techniques for observing folding dynamics, advanced molecular simulations, and artificial intelligence approaches for predicting folding pathways. Particular attention is given to the challenge of observing protein folding in real time, capturing transient intermediate states that determine whether proteins reach their functional structure or misfold.

The event also highlights the interdisciplinary and international nature of the initiative. Representatives from the Embassies of Italy, Japan, and Spain, together with UNESCO, take part in the meeting, emphasizing the global interest in advancing research on protein folding and its biomedical implications. The initiative aims to integrate experimental measurements, computational modeling, and data-driven approaches to build a predictive framework for protein folding dynamics. By combining advanced imaging, force spectroscopy, and machine learning methods, the initiative seeks to better understand how folding processes occur inside living systems and how their disruption can lead to disease.

Overall, the conference provides an opportunity for scientists from different institutions to exchange ideas, establish collaborations, and shape future research directions in the field of protein folding and misfolding. The launch of this initiative represents an important step toward bridging molecular-level observations with biomedical applications, ultimately contributing to improved strategies for diagnosing and treating diseases related to protein misfolding.

Technological Excellence Requires Human and Social Context on ArXiv

Why breakthrough research needs humanities and social sciences. (From an artwork by Jacopo Sacquegno.)
Technological Excellence Requires Human and Social Context
Karl Palmås, Mats Benner, Monica Billger, Ben Clarke, Raimund Feifel, Julia Fernandez-Rodriguez, Anna Foka, Juliette Griffié, Claes Gustafsson, Kerstin Hamilton, Johan Holmén, Kristina Lindström, Tobias Olofsson, Joana B. Pereira, Marisa Ponti, Julia Ravanis, Sviatlana Shashkova, Emma Sparr, Pontus Strimling, Fredrik Höök, Giovanni Volpe
arXiv: 2603.10653

Breakthrough technologies increasingly shape social institutions, economic systems, and political futures. Yet models of research excellence associated with such technologies often prioritize technical performance, scalability, and short-term innovation metrics while treating ethical, social, and cultural dimensions as secondary considerations. This perspective article argues that such separation is no longer tenable. We propose a broader understanding of excellence that combines technical rigor with ethical robustness, social intelligibility, and long-term relevance. The rapid emergence of generative and agentic artificial intelligence further underscores this argument. As technological systems increasingly operate through language, interpretation, and normative alignment, expertise traditionally cultivated in the humanities and social sciences becomes integral to the design, governance, and responsible deployment of such systems. Drawing on historical examples and contemporary research practices, this article examines five interconnected domains where the humanities and social sciences, treated as integrated dimensions of research practice, can strengthen technological development: (1) ethical, legal, and social integration in agenda-setting and research design; (2) plural and reflexive foresight practices that shape technological futures; (3) graduate education as a leverage point for cross-disciplinary literacy; (4) visualization and communication as epistemic and civic practices; and (5) institutional frameworks that move beyond rigid distinctions between basic and applied research. Across these dimensions, we propose practical strategies for embedding interdisciplinary collaboration structurally rather than symbolically.

Roadmap on Deep Learning for Microscopy published in Journal of Physics: Photonics

Spatio-temporal spectrum diagram of microscopy techniques and their applications. (Image by the Authors of the manuscript.)
Roadmap on Deep Learning for Microscopy
Giovanni Volpe, Carolina Wählby, Lei Tian, Michael Hecht, Artur Yakimovich, Kristina Monakhova, Laura Waller, Ivo F. Sbalzarini, Christopher A. Metzler, Mingyang Xie, Kevin Zhang, Isaac C.D. Lenton, Halina Rubinsztein-Dunlop, Daniel Brunner, Bijie Bai, Aydogan Ozcan, Daniel Midtvedt, Hao Wang, Nataša Sladoje, Joakim Lindblad, Jason T. Smith, Marien Ochoa, Margarida Barroso, Xavier Intes, Tong Qiu, Li-Yu Yu, Sixian You, Yongtao Liu, Maxim A. Ziatdinov, Sergei V. Kalinin, Arlo Sheridan, Uri Manor, Elias Nehme, Ofri Goldenberg, Yoav Shechtman, Henrik K. Moberg, Christoph Langhammer, Barbora Špačková, Saga Helgadottir, Benjamin Midtvedt, Aykut Argun, Tobias Thalheim, Frank Cichos, Stefano Bo, Lars Hubatsch, Jesus Pineda, Carlo Manzo, Harshith Bachimanchi, Erik Selander, Antoni Homs-Corbera, Martin Fränzl, Kevin de Haan, Yair Rivenson, Zofia Korczak, Caroline Beck Adiels, Mite Mijalkov, Dániel Veréb, Yu-Wei Chang, Joana B. Pereira, Damian Matuszewski, Gustaf Kylberg, Ida-Maria Sintorn, Juan C. Caicedo, Beth A Cimini, Muyinatu A. Lediju Bell, Bruno M. Saraiva, Guillaume Jacquemet, Ricardo Henriques, Wei Ouyang, Trang Le, Estibaliz Gómez-de-Mariscal, Daniel Sage, Arrate Muñoz-Barrutia, Ebba Josefson Lindqvist, Johanna Bergman
Journal of Physics: Photonics 8, 012501 (2026)
arXiv: 2303.03793
doi: 10.1088/2515-7647/ae0fd1

Through digital imaging, microscopy has evolved from primarily being a means for visual observation of life at the micro- and nano-scale, to a quantitative tool with ever-increasing resolution and throughput. Artificial intelligence, deep neural networks, and machine learning (ML) are all niche terms describing computational methods that have gained a pivotal role in microscopy-based research over the past decade. This Roadmap encompasses key aspects of how ML is applied to microscopy image data, with the aim of gaining scientific knowledge by improved image quality, automated detection, segmentation, classification and tracking of objects, and efficient merging of information from multiple imaging modalities. We aim to give the reader an overview of the key developments and an understanding of possibilities and limitations of ML for microscopy. It will be of interest to a wide cross-disciplinary audience in the physical sciences and life sciences.

Deep-Learning Investigation of Vibrational Raman Spectra for Plant-Stress Analysis on ArXiv

In this work, we present an unsupervised deep learning framework using Variational Autoencoders (VAEs) to decode stress-specific biomolecular fingerprints directly from Raman spectral data across multiple plant species and genotypes. (Image by the Authors of the manuscript. A part of the image was designed using Biorender.com.)
From Spectra to Stress: Unsupervised Deep Learning for Plant Health Monitoring
Anoop C. Patil, Benny Jian Rong Sng, Yu-Wei Chang, Joana B. Pereira, Chua Nam-Hai, Rajani Sarojam, Gajendra Pratap Singh, In-Cheol Jang, and Giovanni Volpe
ArXiv: 2507.15772

Detecting stress in plants is crucial for both open-farm and controlled-environment agriculture. Biomolecules within plants serve as key stress indicators, offering vital markers for continuous health monitoring and early disease detection. Raman spectroscopy provides a powerful, non-invasive means to quantify these biomolecules through their molecular vibrational signatures. However, traditional Raman analysis relies on customized data-processing workflows that require fluorescence background removal and prior identification of Raman peaks of interest-introducing potential biases and inconsistencies. Here, we introduce DIVA (Deep-learning-based Investigation of Vibrational Raman spectra for plant-stress Analysis), a fully automated workflow based on a variational autoencoder. Unlike conventional approaches, DIVA processes native Raman spectra-including fluorescence backgrounds-without manual preprocessing, identifying and quantifying significant spectral features in an unbiased manner. We applied DIVA to detect a range of plant stresses, including abiotic (shading, high light intensity, high temperature) and biotic stressors (bacterial infections). By integrating deep learning with vibrational spectroscopy, DIVA paves the way for AI-driven plant health assessment, fostering more resilient and sustainable agricultural practices.