Ongoing changes in mountain biodiversity have important consequences for the future provision of ecosystem services across scales and for human livelihoods and wellbeing worldwide, calling for effective action. However, the formulation of environmental policies and measures that address the challenges of sustainable management and conservation of mountain ecosystems relies on knowledge that is ‘trapped’ inside a vast and rapidly increasing corpus of unstructured text - the scientific literature, which to date is not accessible to machine-based approaches. Our objective is to develop MoBiKo, an open access global mountain biodiversity knowledge graph built from the entities and relations extracted from the corpus of mountain biodiversity literature. This knowledge graph will ‘liberate’ and structure available knowledge pertaining to the state of, trends in, and drivers of mountain biodiversity. By following principles of findability, accessibility, interoperability, and reusability, we enable broad usage, its expansion with new entities of interest, and its application for varied downstream tasks. Here, we present ongoing work towards achieving a first version of MoBiKo with (i) an approach to improve named-entity recognition based on a hybrid framework that combines structured resources with large language models, and (ii) a preliminary attempt towards relationship extraction using models that are pre-trained on existing datasets and fine-tuned on synthetically generated mountain biodiversity triplets. In addition, we present the domain-specific gazetteers used to address widespread issues of heterogeneous terminologies and enable targeted inference and efficient pre-filtering of relevant sentences, and we provide examples of the contribution of such gazetteers to linked open data and to the systematic mapping of mountain biodiversity literature. Our preliminary results highlight the potential of hybrid and iterative natural language processing pipelines to bridge rule-based and generative methods. By developing this structured, curated, and digitally accessible knowledge base, we aim to support scientific research and inform policy as well as conservation efforts. We further contribute to “opening up what is known about biodiversity” and thereby support the Disentis Roadmap 2024 vision to fully “leverage the power of biodiversity knowledge from research publications within an open science framework”.

Alpine environments are disproportionately affected by anthropogenic climate warming, making them important environments for studying ecosystem responses to rising temperatures. Rising temperatures form feedback loops with soil microbiomes by altering microbial community dynamics, which in-turn impacts greenhouse gas (GHG) emissions and potential further warming. My research aims to identify underlying mechanisms driving microbial community dynamics and GHG fluxes under warming temperatures in the European Alps. More specifically, my research uses controlled laboratory experiments with a synthetic bacterial community to address 1) How does warming impact microbiome composition in alpine ecosystems? 2) How does warming shift the interactions between microbial community members? My work builds upon ongoing field experiments that employ miniature greenhouses, known as “open top chambers”, to locally heat soil at three alpine field sites in the Valais-Wallis region of the Swiss Alps. From these alpine soils, I isolated 12 bacterial strains, each representing a unique genus, to assemble a defined synthetic community. Under controlled laboratory conditions, I grew these strains in monocultures, paired co-cultures, and the full consortia at 10°C and 20°C to simulate a substantial warming scenario. Changes in both relative and absolute abundances will be used to quantify microbe-microbe interactions and assess how their strengths vary with temperature. A knowledge gap exists in studies that link GHG fluxes with genetic profiling of microbial communities in alpine regions, thus my laboratory findings will be also later compared to in-situ field sequencing data of microbial communities and GHG flux measurements. Shifts in microbial community compositions in response to warming will be compared in field and laboratory samples to test for parallel patterns; for example, if similar microbial taxa increase with warmer temperatures in both, this indicates taxa of interest for their GHG metabolisms. Co-occurrences and exclusions between species across temperatures in the natural versus simpler synthetic communities will also be assessed to identify keystone taxa. Understanding how microbial interactions and temperature together impact microbial community composition and associated GHG fluxes will improve predictions of whether alpine ecosystems are likely to act as future sources or sinks of greenhouse gases.

Recently, deep learning approaches to species distribution models (SDMs) have increasingly focused on integrating information-rich modalities such as natural language and remote sensing, motivated by the hypothesis that capturing the non-linear relationships between these inputs and species occurrences should help compensate for limited biodiversity data in poorly monitored regions. However, while leveraging additional modalities has been shown to improve predictions in certain settings, we argue that these improvements remain highly dependent on the task formulation and dataset. We consider the SatBird dataset (Teng et al., 2023) as an illustrative example, showing how leveraging representations derived from satellite imagery does not consistently translate into performance improvements, especially in low-data regimes. We argue that multimodality shouldn't be treated as a generic stepping stone towards improving deep learning-based SDMs, as it can often boil down to the naive assumption that any additional information will be beneficial regardless of their ecological relevance. We also highlight that multimodal approaches in deep learning-based SDMs are predominantly reducible to the inclusion of more and more abiotic covariates, and discuss how such a strategy can amplify the risk of overfitting to sampling biases and amplifying spurious correlations. Finally, we show that leveraging relevant, context-dependent biotic information offers a particularly promising alternative research direction, considering as case studies our work with 1) BATIS (Villeneuve et al., 2026), a novel Bayesian framework that iteratively refines prior predictions from an uncertainty-aware SDM using limited local observations in data-scarce regions, and 2) CISO (Abdelwahed et al., 2025), a novel transformer-based approach that leverages well-documented species groups to improve predictions for data-limited taxa. Results with both BATIS and CISO suggest that universal solutions are unlikely to be sufficient to address current limitations in deep learning-based SDMs, and that further improvements in predictive performance are more likely to come from targeted approaches dedicated to specific data gaps and ecological contexts.

Transcription factors (TFs) regulate gene expression by binding cis-regulatory DNA elements, yet how trans-regulatory characteristics such as TF affinity, concentration, and localization interact with cis-regulatory elements remains largely unclear. We systematically analyzed TF affinity mutants across abundance, and localization states and found that promoter binding-site strength most readily modulated expression levels, followed by TF localization and concentration, while affinity variations were mainly buffered. We further uncover performance trade-offs between TF abundance, localization, and affinity. Together, these results reveal how trans and cis factors collectively shape gene-regulatory output.

Arctic lakes are vital freshwater ecosystems that support wildlife, sustain Indigenous communities, and regulate regional biogeochemical cycles. These lakes experience prolonged periods of snow and ice cover, which shape their physical and chemical structure and profoundly influence microbial life. Rapid climate warming is disrupting these seasonal patterns by altering the duration, thickness, and extent of lake ice, with consequences for underwater light availability, stratification, mixing regimes, and nutrient dynamics. Among the most documented impacts is the progressive shortening of the ice-covered period, yet the biological consequences of these changes—particularly for microbial communities—remain insufficiently explored. Given that bacteria drive key ecological processes in Arctic lakes, including organic matter degradation, primary production, and nutrient cycling, understanding their responses to shifting ice conditions is essential for predicting ecosystem trajectories under climate change. Recent exploratory work in Greenland has revealed that lake ice harbors unexpectedly abundant and metabolically active bacterial communities, often enriched in nitrogen species such as total nitrogen and ammonia compared to underlying waters. These ice-associated assemblages also differ taxonomically and functionally from those in the water column, including exhibiting an enhanced capacity to metabolize complex organic substrates. However, the mechanisms that lead to the development of these distinct microbial communities, their origin (including potential aerosol deposition), and their contributions to nitrogen cycling remain poorly resolved. In this study, we investigated how ice cover structures microbial communities and nitrogen-transforming processes in lakes from both East and West Greenland. We combined environmental monitoring, ice and water chemistry, 16S rRNA gene sequencing, and metabolic assays to characterize microbial assemblages across ice, water column, and associated aerosol particles. Our findings show that lake ice forms a unique habitat enriched in nutrients and selective physicochemical conditions that promote distinct and active bacterial communities. The elevated nitrogen species and metabolic capacities observed in ice suggest that ice cover may act as both a reservoir and a seasonal pulse of microbial biomass and nutrients to the lake during melt. These insights highlight the ecological significance of ice-associated microbiomes and underscore the potential for ongoing shifts in ice phenology to reshape Arctic lake biogeochemistry in a warming climate.

The segmentation clock is a biological oscillator that controls the rhythmic formation of somites during vertebrate embryogenesis. Its core is composed of a transcriptional negative feedback loop in which oscillatory expression of her/Hes genes produces spatially coordinated waves of gene activity in the presomitic mesoderm (PSM). While theoretical models predict that the period of such oscillators should scale with the degradation rate and transcriptional delay of the core repressors, experimental evidence has revealed a more complex landscape, shaped by both intrinsic and extrinsic modulators. This thesis investigates the biophysical and biochemical parameters that influence oscillation timing in the zebrafish segmentation clock, with a particular focus on the protein Her1, a key component of the oscillator.

Using a combination of fluorescent reporter lines, pharmacological perturbations, and single-cell time-lapse imaging, I systematically examined how translation and degradation rates affect Her1 dynamics. Low doses of cycloheximide extended the Her1 oscillation period without altering protein stability, suggesting that translational delay modulates feedback timing. Conversely, proteasome inhibition with MG132 increased Her1 half-life but did not measurably change the period, revealing an apparent decoupling between degradation and timing. This dissociation contradicts the monotonic period-degradation relationship predicted by delay differential equation models and suggests buffering at the network level.

To test whether direct modulation of Her1 degradation can influence timing, I developed a transgenic zebrafish line expressing Her1 fused to a fluorescent tag and an auxin-inducible degron (mAID). In a simplified genetic background lacking functional endogenous clock genes, this construct served as the sole functional component of the core oscillator. Although traveling waves and somite boundaries failed to form in this chassis system, dissociated single PSM cells expressing the transgene retained self-sustained oscillations. These results demonstrate that Her1 alone is sufficient to drive autonomous oscillations, and that this behavior is preserved even in the absence of other core segmentation clock repressors.

Notably, the uninduced mAID tag consistently prolonged the period of Her1 oscillations, across both wild-type and mutant backgrounds. This effect likely reflects intrinsic changes in Her1 feedback kinetics caused by the fusion tag, such as altered folding, dimerization, or DNA-binding efficiency. Furthermore, I found that the oscillation period is highly sensitive to external factors: cells cultured in a low-autofluorescence medium oscillated with significantly longer periods than those in standard L15 medium, likely due to metabolic differences that affect transcription and degradation kinetics.

Together, these findings define Her1 as a minimal genetic oscillator, illuminate how its timing can be modulated by both intrinsic and extrinsic factors, and highlight the resilience of its period to moderate changes in protein turnover. This work underscores the importance of combining theoretical modeling with single-cell assays to uncover the principles that govern developmental timing in dynamic gene regulatory networks.

Mobile genetic elements (MGEs) and bacteria are engaged in an evolutionary arms race, with conjugative plasmids promoting horizontal gene transfer (HGT) and bacterial defence systems restricting their spread. The 7th Pandemic El Tor (7PET) lineage of Vibrio cholerae, responsible for the ongoing cholera pandemic, rarely carry plasmids compared with environmental strains. This low plasmid prevalence is attributed to two 7PET-specific defence systems, DdmABC and DdmDE: DdmABC triggers abortive infection (Abi), causing host cell sacrifice to prevent phage and plasmid replication, whereas DdmDE directly degrades plasmid DNA. Because Abi imposes a substantial burden on cells acquiring large conjugative plasmids, plasmid carriage is strongly selected against in 7PET strains. Unexpectedly, we identified a conjugative plasmid from an environmental V. cholerae strain, pSA7G1, that shows the opposite phenotype. Despite being targeted by both defence systems, pSA7G1 provides a fitness advantage to its 7PET host when competed against plasmid-free cells. This thesis aimed to define the genetic determinants and molecular mechanisms underlying this phenotype. Fluorescence microscopy revealed that plasmid-free cells exhibit rounding, membrane permeabilization and cytoplasmic loss, supporting that pSA7G1-carrying cells outcompete plasmid-free cells through killing. Using serial deletion, we found that both the mating pair formation (MPF) and mobility (mob) modules, typically required for plasmid transfer, are essential for the phenotype, demonstrating that killing is coupled to conjugation. After engineering a two-plasmid system in which an unrelated oriT-containing plasmid is mobilized by co-residing pSA7G1 deleted from its oriT, we observed that transfer of the unrelated plasmid was sufficient for killing. This suggests that donor cells may deliver an effector protein together with transferred DNA. Beyond the conjugation machinery, two additional plasmid regions are required for the phenotype. One contributes to plasmid stability, and its deletion causes plasmid loss and thus abolishes killing. The second contributes directly to the phenotype, as its deletion disrupts killing without altering plasmid stability, host fitness or conjugation. Within the second region, we identified two key genes. One encodes an exclusion protein that blocks pSA7G1 transfer when artificially expressed into plasmid-free cells, thereby protecting them. The second encodes a putative effector protein whose deletion eliminates killing in both pSA7G1 and the two-plasmid system. Although toxicity could not be demonstrated upon induction, the putative effector contains a signal sequence directing its periplasmic localisation, suggesting compartment-based protection in host cells. Together, these findings support a model in which pSA7G1 encodes a novel plasmid addiction system relying on sustained transfer of an effector protein along with plasmid DNA. This mechanism efficiently eliminates cells that fail to establish the plasmid and therefore cannot express the protective exclusion protein, enabling pSA7G1 to overcome the DdmABC and DdmDE defence systems.

Chiral monodentate phosphines are rapidly developing ligand class in asymmetric catalysis. Despite the synthetic utility, full exploration of their application potential remains hampered due to elaborated preparation protocols, especially for scaffolds with axially chiral backbones. Recently, the ortho-directed atroposelective C-H arylation has emerged as efficient tool for atom- and step-economical preparation of atropisomers. Herein we report our results on the synthesis of new family of axially chiral biaryl monophosphines via Ir-catalyzed, phosphine oxide directed C-H arylation as a key step. A large ligand library with a diverse chemical space was successfully prepared, and its application potential was thoroughly evaluated in various benchmark asymmetric reactions. Excellent levels of reactivity and enantioselectivity have been achieved Au(I)-catalyzed cyclisation of enynes and Pd-catalyzed arylative dearomatization of phenols. Moreover, using our new ligand library, an unprecedented atroposelective direct C - H arylation of thiazole derivatives was developed for the first time, providing straightforward access to unique axially chiral N-oxide frameworks. These scaffolds were readily amenable to downstream derivatization and showed promising potential as organocatalysts.

With the deep integration of life science and semiconductor technology, high-sensitivity biosensing and intelligent signal processing have become key directions in next-generation bioelectronics. Biosensing plays a crucial role in early disease detection and cancer risk prediction, offering substantial application potential. Meanwhile, the global biosensor market continues to expand, driven by the increasing need for point-of-care testing, minimally invasive diagnostics, and intelligent medical devices. However, conventional biosensors suffer from complex detection procedures, low integration levels, and poor signal stability, which limit their capability for real-time and high-throughput analysis.

This dissertation introduces a new bio-semiconductor technology framework that unifies semiconductor device physics with interfacial molecular engineering. Experimental investigations reveal that quasi one-dimensional (1D) silicon nanowires (SiNW) exhibit multiple stable and reversible memristive behaviors that can be systematically modulated through interfacial oxidation and controlled surface chemistry. By integrating engineered biofunctionalization pathways, a memristive biosensing system is realized that directly couples molecular recognition processes with intrinsic switching kinetics of the semiconductor device. This approach enables quantitative and label-free detection of small biomolecules while preserving device-level electrical consistency. The proposed architecture supports high-density and multi-channel molecular analysis and provides a compact and manufacturing-compatible route for constructing multifunctional bioelectronic sensors.

Building on this platform, the dissertation further advances neuromorphic biosensing through the dynamic response characteristics of memristive elements. The switching kinetics of the device exhibit strong analogies to synaptic plasticity, allowing biochemical concentration information to be mapped into spiking-like electronic outputs. Network-level demonstrations using memristor arrays show that molecular inputs can be transformed into hardware-encoded signatures, enabling device-level decision making relevant to early cancer risk prediction and biomarker classification. This establishes a rigorous integration pathway from semiconductor physics to neuromorphic network operation, bridging low-level device behavior with high-level functional intelligence.

The demonstrated mechanisms offer new opportunities for implementing autonomous diagnostic nodes capable of real-time molecular sensing without relying on bulky external equipment. Such systems hold significant potential for continuous disease monitoring, early-stage cancer surveillance, and integration into wearable or implantable medical devices. As biosensor markets continue expanding toward intelligent hardware platforms, the technology explored in this dissertation provides an essential foundation for future commercial biosensing systems.

Overall, the findings of this work indicate a promising direction for building fully integrated biosensing architectures in which detection, processing, and decision-making functionalities coexist within a unified semiconductor platform. The demonstrated device behaviors and system configurations advance the long-term vision of hardware-based neuromorphic computation in bio-semiconductor technologies and pave the way for bio-inspired sensing machines

The current paradigm shift in industry, with the advent of the fourth industrial revolution, is driving increasingly fast evolutions in manufacturing technologies. Amidst the rise and abundance of cyber-physical systems have emerged a wide variety of computer-controlled, layer-wise material processing techniques, technically known as Additive Manufacturing (AM) or, more commonly, 3D-printing. Among AM technologies, Powder Bed Fusion - Laser Beam Melting (PBF-LB/M) and powder-based Laser Directed Energy Deposition (L-DED) have consistently and reliably fulfilled certain novel needs of the industry. In both of these processes, a high-energy laser beam is used to locally melt a micrometric metallic powder in a layer-by-layer fashion, leading upon solidification to the production of components with precise and complex geometries. Metallic AM is a complex thermo-physical process that involves many individual, codependent parameters. Fast dynamics, combined with highly localized temperature gradients and extreme cooling rates, give rise to distinctly out-of-equilibrium transformations and the development of important residual stresses. The complex phenomena involved require equally advanced characterization techniques capable of producing comprehensible results through cross-analysis of various sources. In this context, operando (meaning in-process) characterization at synchrotrons and neutron sources is especially relevant, as it allows to directly correlate the structure and properties of a material with its function or its manufacturing process as it operates, revealing complex mechanisms impossible to see otherwise. MADISON-X (Multimodal Advanced DED Investigation System for Operando Neutron and X-ray science) was therefore developed as a platform for operando studies at large-scale facilities during L-DED of most metallic materials. Designed for a scientific community of users, it is capable of multi-material processing, functional grading, and in-situ alloying of any two metallic powders, and was specially engineered to allow operando X-ray tomography during the L-DED process. Its nozzle can also be swapped with a welding torch for Wire Arc Additive Manufacturing (WAAM) operation. Its construction and current capabilities are presented in this thesis. Multimodal advanced characterization is all-the-more relevant in commissioning AM for nuclear applications, e.g. for the production of fusion reactor plasma facing components (PFCs). Complying with nuclear safety regulations requires an outstanding level of understanding and control of a material's microstructure, only complete with the knowledge provided by experiments at large-scale facilities. This work demonstrates how conventional microstructure characterization techniques (such as EBSD mapping) can miss, or misrepresent deeply embedded structures as in the present case, retained austenite formed in the bulk of a PBF-LB/M ferritic-martensitic steel. The application of PBF-LB/M for PFCs manufacturing is investigated, highlighting the formation and microstructure of the steel-tungsten interface, and exploring energy grading as a novel mitigation pathway. Lastly, the application of MADISON to large-scale facility science is showcased, with results from operando X-ray Diffraction studies of the L-DED interface between steel and tungsten. In-situ neutron studies of WAAM demonstrate the machine's versatility and its potential as a community-oriented platform.