Nature has survived billions of years of upheaval—yet coral reefs are collapsing, the Amazon is losing resilience, and lakes can flip into degraded states almost overnight. How can systems this battle-tested also be this brittle? In this talk, I'll argue that the answer lies in a central paradox of complex systems: the "Robust-Yet-Fragile" principle. The very features that confer resilience can also accumulate hidden vulnerabilities to novel or extreme shocks.

I'll walk through a series of empirical examples to show this duality in action, then formalize it through the lens of control theory's "conservation of fragility". The implication is unsettling: managing nature for resilience may be exactly what makes it fragile, and averting collapse requires learning to spot vulnerabilities hiding in plain sight.

Research into the origin of life and its possible presence in the universe has never been more intense. To detect it, specialists from all fields (biologists, geneticists, chemists, physicists and astrobiologists) attempt to define and study its properties: chemical composition, metabolism, reproduction, evolution, self-organisation and complexity. However, whilst small rovers scour the Martian surface in search of it, the most obvious aspect of life on Earth has, paradoxically, never been formally defined: its geometry. Indeed, there is currently no comprehensive inventory of the morphological diversity and complexity of organisms that would enable us to define its fundamental properties. Consequently, essential questions as ‘what kinds of forms can living beings produce?’, ‘how can we recognize life on another planet?’ or ‘are there general limits or rules governing the morphological evolution of life?’ currently have no satisfactory answers at the scale of the biosphere. The reason for this is very simple: there is no theoretical framework that allows us to rationally compare and quantify geometries and structures as complex and diverse as those of the living beings inhabiting our planet.

In this presentation, I will outline a new method based on the structural and fractal properties of shapes, designed to quantify and compare the morphology of all living beings on Earth (e.g., bacteria, ants, whales, sequoias, office colleagues), regardless of their level of complexity and size. By applying this method to all known phyla, we will explore how body size influences the form of life on a large scale and discuss the role of chance and necessity, as well as the constraints and general trends in the evolution of life on Earth (and perhaps in the Universe…).

Cellular populations need to respond collectively to their environmental conditions. A wide range of such collective strategies, evolved over hundreds of millions of years, ensure that multicellular organisms develop correctly to become functional life forms. Yet simpler life forms such as bacteria, which have lived on earth for billions of years, exhibit similar non-trivial responses. In this talk I will discuss some of those strategies.

The human cell nucleus contains roughly two meters of DNA, packed together with proteins that can form biomolecular condensates with different functions such as RNA transcription. In this talk, I will present recent data and theory that active transcription of RNA, a key regulator of condensate formation and dissolution, controls condensate patterning in the nucleolus. Applying this model to transcriptional condensates, and accounting for time delays due to promoter-proximal pausing, leads to spatial oscillations consistent with a “kiss-and-kick” model of interactions between enhancer-bound condensates and promoters. Finally, I will pivot towards condensates in fluctuating viscoelastic fluids and show that coherent motion within a fluid domain will cause pairs of condensates to attract via capillary forces due to induced polarization, providing a potential route for chemically specific coalescence driven by mechanical agitation.

Tumour hypoxia is a defining feature of the tumour microenvironment and a major cause of therapeutic resistance. Although abnormal vascular organisation is widely recognised to impair oxygen delivery, the biophysical mechanisms connecting vascular structure, red blood cell (RBC) transport and tissue hypoxia remain incompletely understood. In this talk, I will summarise published work showing how specific vascular abnormalities disrupt RBC partitioning and thereby generate heterogeneous oxygenation. Quantitative analysis of tumour vascular networks, combined with mathematical modelling, showed that a reduced vessel length-to-diameter ratio biases haematocrit distribution and promotes spatial heterogeneity in oxygenation, identifying this metric as a structural surrogate of tumour perfusion and oxygenation (Bernabeu et al., PNAS 2020). Complementary computational studies further showed that vessel compression first perturbs RBC partitioning at individual bifurcations (Enjalbert et al., PNAS 2021), before these local disturbances propagate across networks and increase haematocrit heterogeneity (Enjalbert et al., Communications Physics 2024). Together, these studies establish a mechanistic framework linking abnormal vascular structure, disordered RBC transport and tumour hypoxia.

I will then present unpublished data from a glioblastoma (GBM) model that test this framework experimentally. In GBM, focal adhesion kinase (FAK) has been implicated in tumour progression, invasive behaviour and tumour–microenvironment interactions. To examine whether tumour cell FAK also shapes the vascular determinants of hypoxia, we used a tumour cell FAK knockout model. Loss of FAK increased tumour hypoxia without altering intrinsic tumour cell oxygen consumption, and this effect was not explained by vascular density. Instead, FAK-deficient tumours exhibited more abnormal three-dimensional vascular architecture. We further show that the vessel length-to-diameter ratio is the strongest microvascular predictor of hypoxic fraction, with lower values associated with greater hypoxia. These results suggest that tumour-cell-driven abnormalities in vascular structure, rather than vessel abundance alone, are closely associated with impaired oxygenation in GBM.