Yesterday, October 9, 2026, Jing Yang successfully defended her PhD thesis, “A Quantitative View of the Heterogeneity–Diversity Axis in Biological Systems — From the Tumour Microenvironment to the Pond Ecosystem,” in the Sala Segimon Serrallonga at UVic-UCC, within the PhD programme in Bioinformatics.

Jing’s thesis has been interdisciplinary from start to finish. It looks at tumours and ecosystems, two systems that seem very different, through a common question: how diversity and heterogeneity among their parts shape stability, adaptation and change. On the biomedical side, she used single-cell transcriptomics to study cellular plasticity and therapeutic resistance in glioblastoma, and to describe the transition between non-muscle-invasive and muscle-invasive bladder cancer. She combined dimensionality reduction, network inference and perturbation analyses to identify key cell-state transitions and regulatory hubs. On the ecological side, she analysed field data from 240 ponds in 41 pondscapes across Europe to work out how climate, land use and pond features drive nitrogen and phosphorus enrichment.

Although the two systems are studied separately, the thesis closes with a reflection on what they may have in common. It discusses how concepts such as entropy and ecological diversity metrics could be applied to both cancer and ponds. Rather than a finished integration, this opens the door to a shared quantitative language for resilience and transformation in complex biological systems, a line we hope to keep exploring between oncology and ecology.

This thesis is also a milestone for the CBBL. It is the first thesis in our group devoted to the analysis of omics data, and the first carried out in collaboration with the Aquatic Ecology Group of the Faculty of Sciences, Technology and Engineering (FCTE), as part of the PRISTIN project led by Mireia Bartrons. The thesis was supervised by Jordi Villà-Freixa, Adrián López García de Lomana and Mireia Bartrons.

We would like to warmly thank the members of the examining committee for their careful reading of the thesis and for an engaging discussion:

• Dr. Marc Llirós Dupré (UVic-UCC), Chair
• Dr. Miquel De Cáceres Ainsa (CSIC–CREAF), Secretary
• Dr. Carolina Trochine (INIBIOMA, CONICET–Universidad Nacional del Comahue), Member

Our thanks also go to the substitute members, Dr. Thomas Duflot (Université de Rouen), Dr. Mireia Olivella García (UVic-UCC) and Dr. Ottar Rolfsson (University of Iceland), and to everyone who joined us for the defence.

Congratulations, Dr. Yang, and all the best for the next steps!

Our student Bernat Casajuana presented a poster at the 3rd Biology for Physics Conference, held September 6 to 10, 2026, at the Barcelona Biomedical Research Park, under the theme “Energy and Information in Living Matter.” The conference combined invited talks, contributed talks, and poster sessions, with a particular focus on connecting early-career researchers with established leaders in the field.

The conference explored living systems through nonequilibrium statistical physics, focusing on how biological function depends on energy dissipation and information processing far from equilibrium. Core concepts included entropy production as a signature of irreversibility, thermodynamic and information-theoretic limits on cellular sensing and signaling, and methods for inferring physical laws directly from noisy experimental data.

That last topic is close to what Bernat has been working on with physics-informed neural networks (PINNs), where instead of learning the physical laws from data, the laws are already known and are used to guide a machine-learning model for parameter recovery in the Repressilator, a small synthetic gene network.

Work carried out with Roger Casals-Franch, Adrián López García de Lomana, Pere Martí-Puig, and Jordi Villà-Freixa (IRIS-CC · UVic-UCC · University of Iceland).

Bernat had presented the project a few days earlier, as a talk, at MDAI 2026 in Vic.

Bernat Casajuana, an undergrad Biomedical student at the CBBL, attended and presented at MDAI 2026, the 23rd International Conference on Modeling Decisions for Artificial Intelligence, held in Vic from September 7 to 9, 2026.

The meeting focused on the mathematical foundations of decision-making in AI, including the aggregation of information from multiple sources, human decision models, and explainable machine learning—concepts close to how physics-informed neural networks combine data and physical priors in a single model.

Bernat’s talk, “Physics-Informed Neural Networks for Parameter Recovery in the Repressilator Oscillatory Model,” presented work on using PINNs to recover the true kinetic parameters of the Repressilator, a small synthetic gene oscillator, directly from noisy and partially observed simulated data.

Work carried out with Roger Casals-Franch, Adrián López García de Lomana, Pere Martí-Puig, and Jordi Villà-Freixa (IRIS-CC · UVic-UCC · University of Iceland).

Bernat presented his project a few days later, as a poster, at the 3rd Biology for Physics Conference in Barcelona.

Last week, lab members Roger Casals and Ian Carbó, together with Mireia Navarro and Gabriel Ruiz, took part in the Respira Hackathon, held in person from October 1 to 3, 2026, at AstraZeneca’s Estel building (Av. Roma 81, Barcelona), where their project received the award for the COVID-19 challenge. Both Roger and Gabriel are PhD candidates in the interuniversity PhD programme in Bioinformatics at UVic-UCC: Roger is supervised by Jordi Villà-Freixa and Adrián López García de Lomana, and Gabriel by Mireia Olivella.

Respira was the first international CIBER hackathon on respiratory diseases. It was jointly organized by CIBERES, the Spanish biomedical research network on respiratory diseases (part of the Instituto de Salud Carlos III and the Spanish Ministry of Science, Innovation and Universities), and AstraZeneca. The Facultat d’Informàtica de Barcelona (UPC) and the Barcelona Supercomputing Center (BSC-CNS) collaborated in the organization, providing their expertise and advanced computing capacity, with further support from the Universitat de Barcelona and Kunsen. For 48 hours, teams of up to four people with diverse backgrounds worked on real clinical problems, guided by experts in respiratory medicine, genetics, immunology and bioinformatics, before presenting their solutions to a jury.

The event proposed four challenges, each framed from both a health and a technology perspective:

• Pulmonary fibrosis: finding a simple molecular signature, measured at the diagnosis of interstitial lung abnormalities, that identifies the patients most likely to progress to idiopathic pulmonary fibrosis, using clinical variables and serum levels of around 3,000 proteins.
• COPD: automatically extracting subtle structural changes from CT scans and combining them with clinical and molecular data to detect early lung damage before the disease can be diagnosed by spirometry.
• Severe asthma: designing an AI-based “digital twin” of the patient to simulate treatment strategies with biologic therapies and anticipate their outcomes.
• COVID-19 and other respiratory viruses: turning clinical, functional, radiological and follow-up data into a map of patient profiles that recover in similar ways after a severe respiratory infection.

Our team took on the COVID-19 challenge, which focused on the long-term after-effects of COVID-19 and other severe hospital stays, including intensive care (ICU), using patient datasets collected by different hospitals. After a severe respiratory infection, some patients recover fully, while others suffer for months or years from breathlessness, fatigue, memory problems or muscle weakness. Current classifications often lump together people with very different needs, so finding reproducible, clinically meaningful patient profiles is a first step towards more personalized care.

Drawing on their multidisciplinary expertise, the team developed a solution that connects data science and clinical practice, a link between the two fields that is both urgent and much needed. Their proposal, “Early Prediction of Post-COVID-19 Clinical Follow-Up Need: A Tool for Clinicians,” is a two-stage hierarchical Bayesian model that accounts for differences between hospitals. It was validated on later patients and on hospitals the model had never seen.

The jury awarded the team first prize in the COVID-19 challenge, shared ex aequo with the team “ResFen.”

Congratulations to the whole team and the rest of participants!