In the current era of the Anthropocene, human activities are powerful forces that affect the geosphere, atmosphere, and biosphere – globally, fundamentally, and in most cases irreversibly. In freshwaters, land use change, chemical pollution, decline in biodiversity, spread of invasive species, climate change, and shifts in the hydrological regime are among the key drivers of changes. In the 21st century, major water engineering projects such as large dams and water diversion schemes will fundamentally alter the natural hydrological regime of entire landscapes and even continents. At the same time, the hydrological regime is the governing variable for biodiversity, ecosystem functions and services in river networks. Indeed, there will be an increasing conflict between managing water as a resource for human use and waters as highly valuable ecosystems. Therefore, research needs to unravel the challenges that the freshwaters are facing, understand their potential drivers and impacts, and develop sustainable management practices – for the benefit of humans and ecosystems alike. The present thesis focuses on three currently understudied alterations in flow and material dynamics within river networks, namely (i) on the dynamics of floating organic matter (FOM) and its modification in dammed rivers, (ii) on river intermittency and its effects on nutrient and organic matter (OM) dynamics, and (iii) on major future water transfer schemes. Massive construction and operation of dams cause modification of water flow and material fluxes in rivers, such as of FOM. FOM serves as an essential component of river integrity, but a comprehensive understanding of its dynamics is still lacking. River damming, climate change and water extraction for human needs lead to a rapid expansion in number and extent of intermittent rivers worldwide, with major biogeochemical consequences on both regional and global scales. Increased intermittency of river networks also forces people to implement engineering solutions, such as water transfer schemes, which help to supply water to places of demand. Water transfer projects introduce artificial links among freshwater bodies modifying the hydrological balance. Impacts of abovementioned activities on freshwaters have been assessed in single case studies. However, the current knowledge does not allow a generalization of their globally applicable meaning for ecosystems. Furthermore, mostly neglected aspects of these alterations, such as the potential consequences of FOM extraction from rivers, the biogeochemical role of intermittent rivers upon rewetting, and the current scale of water transfers require better understanding before bold conclusions could be made. By combining research methods such as extensive literature reviews, laboratory experiments and quantitative analyses including spatial analyses with Geographic Information Systems, I investigated (1) the natural cycle, functions, and amounts of FOM in rivers fragmented by dams, (2) effects of rewetting events on the pulsed release of nutrients and OM in intermittent rivers and ephemeral streams (IRES), and (3) the potential extent of water transfer megaprojects (WTMP) that are currently under construction or in the planning phase and their role in modifying the global freshwater landscape. In all three cases, I provide a global perspective. The role of FOM in rivers as a geomorphological agent, a resource, a dispersal vector and a biogeochemical component was investigated based on an extensive literature review. Collected information allowed for conceptualizing its natural cycle and dynamics, applicable to a wide range of rivers. Data on FOM accumulations at 31 dams located within catchments of 13 rivers showed that damming leads to FOM entrapment (partly or completely) and modifies its natural cycling. The results of a spatial analysis considering environmental properties revealed that catchment characteristics can explain around 57% in the variation of amounts of trapped FOM. Effects of rewetting events on the release of nutrients and OM from bed sediments and course particulate organic materials (CPOM) accumulated in IRES was studied in laboratory experiments. Using a large set of samples collected from 205 rivers, located in 27 countries and distributed across five major climate zones, I determined the concentrations and qualitative characteristics of nutrients and OM released from sediments and CPOM. I also assessed how these characteristics can be predicted based on environmental variables within sampled IRES. In addition, I calculated area-specific fluxes of nutrients and OM from dry river beds. I found that the characteristics of released substances are climate specific. In the Continental zone I found the highest concentrations of released nutrients, but the lowest quality of OM in terms of its potential bioavailability. In contrast, in the Arid zone the concentrations of released nutrients were the lowest, but the quality of OM the highest. The effect of environmental variables on the concentrations of nutrients and the quality of OM was better predicted for sediments than for other substrates with the highest share of explained variance in the Continental and Tropical zones. On the global scale, dissolved organic carbon, phenolics, and nitrate dominate fluxes released during rewetting events. Overall, this study emphasized that on the global scale rewetting events in IRES represent biogeochemical “hot moments†, but characteristics of released nutrients and OM differ greatly among climate zones. The present thesis fills also a major knowledge gap on the global distribution of large water transfer schemes (referred to as “megaprojects†) that are actually planned or under construction. To provide an inventory of WTMP, I collected data from various literature sources, ranging from published academic studies, the official web-sites of water transfer projects, environmental impact assessments, reports of non-governmental organizations, and information available in on-line newspapers. In total, 60 WTMP were identified. Information on spatial location, distances and volumes of water transfer, costs, and purposes of WTMP was collected and compared with those of existing schemes. The results showed that North America, Asia and Africa will be the most affected by future WTMP having the highest densities of projects and the largest water transfer distances and volumes. If all projects were completed by 2050, the total water transfer distances would reach 77,063 km transferring more than 1,249 km3 per year, which corresponds to about 20 times the annual flow of the river Rhine. The outcomes of the thesis provide major implications for environmental management. Natural FOM is an important component for sustaining the ecological and geomorphic integrity of rivers and, therefore, should be managed appropriately. Intermittent rivers must be considered in models quantifying nutrient and OM fluxes in river networks. First flush events in particular release huge amounts of nutrients and OM, which may cause dramatic metabolic effects on downstream receiving waters. Finally, the future WTMP alter the hydrological balance of entire river basins and continents. They require multiple assessments before construction and careful management practices for sustainable operation in order to consider both freshwater as a resource as well as freshwaters as pivotal ecosystems.

This thesis develops an Artificial Intelligence (AI) approach intended for accurate patient stratification and precise diagnostics/prognostics in clinical and preclinical applications. The rapid advance in high throughput technologies and bioinformatics tools is still far from linking precisely the genome-phenotype interactions with the biological mechanisms that underlie pathophysiological conditions. In practice, the incomplete knowledge on individual heterogeneity in complex diseases keeps forcing clinicians to settle for surrogate endpoints and therapies based on a generic one-size-fits-all approach. The working hypothesis is that AI can add new tools to elaborate and integrate together in new features or structures the rich information now available from high-throughput omics and bioimaging data, and that such re- structured information can be applied through predictive models for the precision medicine paradigm, thus favoring the creation of safer tailored treatments for specific patient subgroups. The computational techniques in this thesis are based on the combination of dimensionality reduction methods with Deep Learning (DL) architectures to learn meaningful transformations between the input and the predictive endpoint space. The rationale is that such transformations can introduce intermediate spaces offering more succinct representations, where data from different sources are summarized. The research goal was attacked at increasing levels of complexity, starting from single input modalities (omics and bioimaging of different types and scales), to their multimodal integration. The approach also deals with the key challenges for machine learning (ML) on biomedical data, i.e. reproducibility, stability, and interpretability of the models. Along this path, the thesis contribution is thus the development of a set of specialized AI models and a core framework of three tools of general applicability: i. A Data Analysis Plan (DAP) for model selection and evaluation of classifiers on omics and imaging data to avoid selection bias. ii. The histolab Python package that standardizes the reproducible pre-processing of Whole Slide Images (WSIs), supported by automated testing and easily integrable in DL pipelines for Digital Pathology. iii. Unsupervised and dimensionality reduction techniques based on the UMAP and TDA frameworks for patient subtyping. The framework has been successfully applied on public as well as original data in precision oncology and predictive toxicology. In the clinical setting, this thesis has developed1: 1. (DAPPER) A deep learning framework for evaluation of predictive models in Digital Pathology that controls for selection bias through properly designed data partitioning schemes. 2. (RADLER) A unified deep learning framework that combines radiomics fea- tures and imaging on PET-CT images for prognostic biomarker development in head and neck squamous cell carcinoma. The mixed deep learning/radiomics approach is more accurate than using only one feature type. 3. An ML framework for automated quantification tumor infiltrating lymphocytes (TILs) in onco-immunology, validated on original pathology Neuroblastoma data of the Bambino Gesu’ Children’s Hospital, with high agreement with trained pathologists. The network-based INF pipeline, which applies machine learning models over the combination of multiple omics layers, also providing compact biomarker signatures. INF was validated on three TCGA oncogenomic datasets. In the preclinical setting the framework has been applied for: 1. Deep and machine learning algorithms to predict DILI status from gene expression (GE) data derived from cancer cell lines on the CMap Drug Safety dataset. 2. (ML4TOX) Deep Learning and Support Vector Machine models to predict potential endocrine disruption of environmental chemicals on the CERAPP dataset. 3. (PathologAI) A deep learning pipeline combining generative and convolutional models for preclinical digital pathology. Developed as an internal project within the FDA/NCTR AIRForce initiative and applied to predict necrosis on images from the TG-GATEs project, PathologAI aims to improve accuracy and reduce labor in the identification of lesions in predictive toxicology. Furthermore, GE microarray data were integrated with histology features in a unified multi-modal scheme combining imaging and omics data. The solutions were developed in collaboration with domain experts and considered promising for application.

Intracellular chloride and pH are fundamental regulators of neuronal excitability and they are often co-modulated during excitation-inhibition activity. The study of their homeostasis requires simultaneous measurements in vivo in multiple neurons. Combining random mutagenesis screening, protein engineering and two-photon-imaging this thesis work led to the discovery of new chloride-sensitive GFP mutants and to the establishment of ratiometric imaging procedures for the quantitative combined imaging of intraneuronal pH and chloride. These achievements have been demonstrated in vivo in the mouse cortex, in real-time monitoring the dynamic changes of ions concentrations during epileptic-like discharges, and in glioblastoma primary cells, measuring osmotic swelling responses to various drugs treatment.

MicroRNAs are a class of small non-coding RNAs involved in post-transcriptional regulation. Their role in almost all processes of the cell, make microRNAs ubiquitary players of cell development, growth, differentiation, cell to cell communication and cell death. Thus, cells’ physiological or pathological conditions are reflected by variations in the levels of expression of microRNAs, enabling them to be used as biomarkers of such states. In the past decade, there has been an exponential increase of studies using microRNAs as potential biomarkers for cancer, neurodegenerative diseases, inflammation and cardiac diseases, from tissues and liquid biopsies. However, none of them has reached the clinics yet, due to inconsistency of results through the literature and lack of assay standardization and reproducibility. Technological limitations of microRNAs detection have been, to date, the biggest challenge for using these molecules in clinical settings. In fact, although microarrays, RT-qPCR and RNA-seq are well-established technologies, they all require complex procedures and trained personnel, for performing RNA extraction, labelling of the target and PCR amplification. All these steps introduce variability and, in addition, since no universally standardized protocol – from sample extraction to analyte detection - has been produced yet, methodological procedures are difficult to reproduce. For this reason, we developed a new platform for the rapid detection of microRNAs in biofluids composed of an innovative silicon-photomultiplier (SiPM) based detector and a new chemistry for nucleic acid testing (Chem-NAT). Chem-NAT exploits a dynamic labelling chemistry which allows the sensitive detection of nucleic acids till single base level. On the other hand, SiPM-based device, compared to normal vacuum photomultipliers, grants miniaturization and higher capacity of fitting in a bench-top solution for clinical settings, among other advantages. The new platform – ODG – has been validated for the direct detection – neither RNA extraction nor PCR amplification needed - of microRNA-21 in plasma of lung cancer patients. In this work, we also explored the use of microRNAs as biomarkers in metastatic castration resistant prostate cancer (mCRPC). We collected plasma samples from mCRPC patients before and after abiraterone acetate treatment – androgen deprivation type of drug – and performed a miRnome analysis for discovering microRNAs predicting the efficacy of the drug. We chose miR-103a-3p and miR-378a-5p and we validated them via TaqMan RT-qPCR. We discovered that the ratio between the two microRNAs is able to predict the efficacy of abiraterone acetate and follow the responsiveness in time. In liquid biopsies, extracellular vesicles are getting increasing importance for diagnostic and prognostic purposes. Therefore, in this work we also explored the expression of some microRNAs in extracellular vesicles from plasma, isolated via nickel-based method. We discovered that microRNA-21 and microRNA-223 are not enriched in vesicles from healthy individuals.