IRCCS San Raffaele Scientific Institute Showcase

This dataset refers to a study where 558 patients with multiple sclerosis (MS) and 214 healthy controls (HC) underwent neurological assessment and 3T MRI acquisition, including resting-state functional MRI (RS fMRI). In addition, 491 MS patients completed a comprehensive neuropsychological evaluation for cognitive status assessment. RS functional connectivity (FC) abnormalities were quantified using degree centrality analysis to characterize large-scale brain network reorganization associated with MS. Spatial patterns of functional abnormalities were then compared between HC and different MS phenotypes, including relapsing-remitting and progressive MS, as well as between cognitively preserved and cognitively impaired patients. To investigate the biological basis of RS FC, spatial correlations were performed between regional degree centrality abnormalities and the expression of 3634 MS-related genes derived from the Allen Human Brain Atlas (AHBA). Genes significantly associated with imaging-derived network alterations underwent pathway enrichment analysis. Compared to HC, MS patients showed increased degree centrality mainly in the default-mode network (DMN), associated with genes involved in inflammation resolution and immune regulation, and reduced centrality in salience network and cerebellar regions linked to cytokine-response genes. Progressive MS patients exhibited higher centrality in DMN and cerebellar areas than HC and relapsing-remitting MS patients, correlating with genes related to epigenetic and mitochondrial functions. Among MS patients, 144 (29.3%) were cognitively impaired and showed increased centrality in DMN and mesial temporal regions compared to cognitively preserved patients and HC. These abnormalities negatively correlated with DNASE1 and CP gene expression, implicated in DNA degradation, iron homeostasis, and neurodegeneration. Regional gene expression spatially correlated with MS-related functional network abnormalities.

Spatial regulation of mRNA polyadenylation is a key emerging mechanism shaping cellular function. The TENT5/FAM46 family comprises four non-canonical poly(A) polymerases that stabilize transcripts encoding ER-targeted proteins, with mutations linked to diseases of professional secretory cells. Using transcriptomic profiling coupled with systematic mutagenesis, we uncover how paralog-specific evolutionary divergence dictates distinct subcellular localization, interaction, and functional outcomes. TENT5D, the most ER-associated member, triggers widespread proteome remodeling, boosting the concerted expression of ER, ERGIC, Golgi, and lysosomal proteins. In contrast, TENT5B lacks ER targeting and regulates proteins involved in cell division. Mechanistically, a member-specific C-terminal region that binds ER-transmembrane FNDC3 proteins is necessary and sufficient for ER localization. Altogether, our findings reveal a domain-encoded mechanism linking TENT5 localization to transcript selectivity and secretory output. The TENT5–FNDC3 axis thus emerges as a key orchestrator of the cellular translatome and organelle identity. OMICS Data: proteomics, RNA-seq and TAIL-Seq Proteomics data folder contains raw data. RNA-seq folder contains BAM files while Fastq are available at GSE326742 (related link below) TAIL-seq folder contains polyA tail lenght results for each samples while Fastq files are available at GSE326755 (related link below)

The EVASION study was a prospective, observational, proof-of-concept study designed to quantify and characterize the HIV reservoir in people living with HIV (PLWH) on suppressive antiretroviral therapy with active oncogenic viral co-infections (HCV, EBV, HPV) and/or virus-driven malignancies, compared to co-infection-free PLWH. In the published work, the EVASION participants with active viral co-infections serve as the control group. The table and figures presented here refer to the entire EVASION cohort and are therefore not restricted to the control group included in the publication.

This dataset was generated to investigate whether integrating demographic, clinical, and magnetic resonance imaging (MRI) features can improve the classification of multiple sclerosis (MS) patients, distinguish disease phenotypes, and predict disability severity. The underlying research hypothesis is that machine learning (ML) models trained on multimodal data can improve disease characterization more accurately than traditional statistical approaches. The dataset includes data from 1554 patients with MS and 520 healthy controls (HC), collected within the the Italian Neuroimaging Network Initiative. For each participant, demographic information (e.g., age, sex), clinical assessments, and brain MRI scans were acquired. Clinical disability was quantified using the Expanded Disability Status Scale (EDSS) score. MRI acquisition included T2-weighted and 3D T1-weighted sequences, from which quantitative imaging features were derived. These features included total and regional T2 lesion volumes (LV), as well as normalized volumetric measures of cortical and subcortical grey matter (GM), white matter, cerebellum, and brainstem. All imaging-derived variables were preprocessed and harmonized across sites. ML models applied to these data (including support vector machines, multi-layer perceptron networks, Random Forest, and Gradient Boosting) demonstrated high performance in disease and phenotype classification, as well as in EDSS prediction. Specifically, classification accuracy for MS vs HC ranged from 89% to 96%, while phenotype classification reached approximately 92% accuracy. Disability prediction achieved strong agreement with observed EDSS scores (intra-class correlation coefficients between 0.7456 and 0.76). Feature importance analyses (using SHAP values) indicated that T2 LV and regional GM volumes (particularly in the brainstem, cerebellum, thalamus, and cortex) were among the most influential variables for classification and prediction tasks. Demographic variables such as age and sex, along with clinical disability scores, also contribute significantly to model performances. This dataset provides a robust, multimodal resource for advancing ML approaches to MS classification and prognosis.

The present includes the raw data from the article "The Impact of Aging and Sex on Corneal Nerve Density and Function" (doi:https://doi.org/10.1167/iovs.67.3.3). In this work, we described the role of sex in modulating mouse corneal nerve morphology and function during aging.

Rationale: to evaluate Fibroblast Growth Factor 21 (FGF21) as a possible biomarkers for depressive symtopms in elderly. FGF21 is a stress-induced hepatokine involved in inflammation and neuroendocrine regulation, processes implicated in depression. Methods: We investigated the association between FGF21 and depressive symptoms in community-dwelling older adults. Data were obtained from 52 older individuals (median age 70; 61.5% women) within the FRASNET cohort who underwent longitudinal assessments (2017 for baseline –2024 for follow up). Depressive symptoms were evaluated using the 15-item Geriatric Depression Scale (GDS). Regression models adjusted for age, sex, and body mass index were used to assess associations between FGF21 and depressive symptoms. Results: Circulating FGF21 levels declined significantly over time (p = 0.03) but remained higher in individuals with depressive symptoms at both baseline and follow-up. Elevated baseline FGF21 predicted higher GDS scores at follow-up (adjusted B = 0.003, 95% CI 0.000–0.006, p = 0.049). Discriminatory performance for elevated depressive symptoms was modest (AUC = 0.66). Conclusions: Higher baseline FGF21 levels were associated with greater depressive symptom burden at follow-up. These findings should be considered preliminary and hypothesis-generating. Further studies using diagnostic outcomes and larger samples are warranted.

Rationale: To identify circulating biomarkers of fraily. Participants were eligible if they were ≥65 years old, able to walk >500 m without assistance, and had a life expectancy >6 months. Exclusion criteria included severe cognitive impairment (Mini-Mental State Examination [MMSE] <18), inability to provide informed consent, or severe health conditions such as uncontrolled hypertension, recent fractures of the upper or lower extremities, or myocardial infarction within the previous year. For this analysis, individuals recruited in nursing homes and those with missing data required to compute frailty or body composition were excluded. Data were collected at baseline (2017) and follow-up (2024). Frailty was assessed using three instruments: the Frailty Index (FI), the modified Frailty Phenotype (FP), and the Clinical Frailty Scale (CFS). Nutritional status was evaluated with the Mini Nutritional Assessment–Short Form (MNA-SF) and categorized as malnourished (≤7), at risk (8–11), or well-nourished (12–14). Sarcopenia was screened using the SARC-F questionnaire and confirmed according to EWGSOP2 criteria. Muscle strength was assessed using the SPPB chair-stand test (>15 s indicating reduced strength). Muscle mass was measured by bioelectrical impedance analysis (BIA), with cut-offs <32.9 % in men and <23.9 % in women. Sarcopenic obesity was defined according to ESPEN/EASO criteria as the coexistence of reduced muscle strength, low muscle mass, and excess fat mass (≥30 % in men and ≥42 % in women). Physical activity was assessed with the Physical Activity Scale for the Elderly (PASE; <76 indicating low activity). Cognitive function was measured using the MMSE (≤24 indicating impairment), and depressive symptoms using the Geriatric Depression Scale (GDS-15). Functional status was evaluated through Activities of Daily Living (ADL) and Instrumental ADL (IADL), and fall risk using the Tinetti scale. Biomarker analyses were performed in plasma samples from 52 of 228 participants due to assay costs and sample availability. Participants were randomly selected among those reassessed at follow-up and were representative of the overall cohort. Blood samples were collected in EDTA tubes from non-fasting subjects, processed within 24 h, aliquoted, and stored at −80 °C. Biomarker concentrations were measured using the ELLA™ automated immunoassay system (Bio-Techne, USA). Continuous variables were reported as mean ± SD or median (IQR), and categorical variables as frequencies (%). Baseline and follow-up comparisons used paired t-tests or Wilcoxon tests. Associations between biomarkers and clinical outcomes were examined using regression models adjusted for age and sex. GDF-15 and FGF-21 concentrations were log-transformed. ROC analyses evaluated the ability of baseline biomarkers to discriminate adverse outcomes at follow-up. Optimal thresholds were determined using the Youden index. Analyses were performed using IBM SPSS Statistics version 25.

Rationale: to identify circulating biomarkers for malnutrition in elderly. Background: Malnutrition is a prevalent geriatric syndrome, with multifactorial etiology and consequences for health and independence. Inflammation contributes to nutritional decline, yet conventional inflammatory markers often lack sensitivity for identifying malnutrition risk. The soluble receptor for advanced glycation end-products (sRAGE), a modulator of inflammatory responses, has emerged as a biomarker of disease risk and adverse outcomes in various conditions. Objectives: to evaluate the association between circulating sRAGE levels and nutritional status in community-dwelling older adults. Methods: This prospective observational study was conducted within the FRASNET cohort. Fifty-two community-dwelling older adults underwent multidimensional geriatric assessments during two time periods: 2017–2020 and 2023–2024. Serum sRAGE levels were measured at both timepoints. Nutritional status was assessed using the Mini Nutritional Assessment Short Form (MNA-SF). Associations between sRAGE and clinical parameters were evaluated through linear regression models adjusted for age and sex. The diagnostic performance of sRAGE in identifying malnutrition was assessed using ROC curve analysis. Results: Higher baseline sRAGE levels were significantly associated with lower BMI (β = –0.003, p = 0.036), reduced calf circumference (β= –0.002, p = 0.04), and poorer nutritional status as revealed by MNA-SF scores (β= –0.001, p = 0.03) at follow-up. Associations were more pronounced in women. ROC analysis indicated good diagnostic accuracy for identifying malnutrition risk, with an AUC of 0.85. The optimal sRAGE cut-off value for malnutrition risk was 1,362.5 pg/mL. Conclusions: Higher sRAGE levels were prospectively associated with poorer nutritional outcomes in older adults, particularly in women. sRAGE may aid early identification of inflammation-related malnutrition risk.

Machine Learning (ML) applied to healthcare Real World Data (RWD) may improve patient management. RWD, however, requires extensive preprocessing to make it ML-ready. The aim of this resource is to explore the impact of preprocessing on ML models applied to RWD from type 2 diabetes patients visits. A GitHub repo is available with the code related to the experiments, to show all the steps we prepared, the decisions we took and alternative pipelines we created. Work done in partial fulfillment of my PhD dissertation: "Leveraging real-world data with machine learning to disentangle the complexity of multimorbid internal medicine patients" Linked in the paper: Montagna M, Rabadzhiev AS, Traverso A, Setola E, Draetta E, Dimonte A, Barbieri S, Fabiani B, Piemonti L, Esposito A, Tacchetti C and Rovere Querini P (2026) From raw data to actionable insights: preprocessing real-world data for machine learning in diabetes care. Front. Digit. Health 8:1685842. doi: 10.3389/fdgth.2026.1685842

Preterm (PRET) neonates are characterized by reduced mature hematopoietic cell count and increased risk of infection, but the biological features of their Hematopoietic stem/progenitor cells (HSPC) compartment, responsible to produce immune cells, are largely unknown. Here, we exploited an optimized workflow to evaluate the immunophenotype, clonogenic potential and differentiation properties of peripheral blood (PB) HSPC from 35 PRET (< 32 weeks gestation) and 26 TERM neonates at birth. Compared to TERM, PRET displayed increased circulation of HSPC, Hematopoietic Stem Cells (HSC) and Multi-Potent Progenitors (MPP). Functionally, TERM and PRET HSPC showed a comparable clonogenic potential, although with a skew towards erythroid colonies of PRET precursors, and similar multi-lineage production. Strikingly, PRET-HSC+MPP and TERM-HSC+MPP displayed equivalent differentiation efficiency and were respectively enriched in uni-erythroid, in line with the erythroid-biased fetal hematopoiesis within the liver, and in uni-lymphoid clones, possibly reflecting the BM output. Finally, PRET with subsequent late-onset neonatal sepsis, who were even more premature, had lower counts of mature lymphoid and myeloid cells but not of HSPC subpopulations. Altogether, PRET- and TERM-HSPC showed comparable in vitro clonogenic and differentiation efficiency. In summary, the functional properties of neonatal HSPCs evolve with the gestational age, from more erythroid-biased fate at earlier stages towards a more myeloid/lymphoid-skewed composition due to distinct hematopoietic niche factors acting on HSPCs with similar differentiation properties. Moreover, our data support the hypothesis that a higher degree of immaturity of the hematopoietic system might increase the susceptibility to develop septic events in preterm newborns.