Poster contributions

Abstract: 

Abstract: With new target therapies and immunotherapies in oncology, the dose-efficacy relationship is no longer clearly increasing. Then, the choice of the maximum tolerated dose for further development is called into question from the scientific community and health authorities. Therefore, the dose recommendation in early-phase oncology trials should not be only guided and oriented by the occurrence of Dose Limiting Toxicities (DLTs), but also by preliminary signs of clinical or biological activity of the experimental treatment 

Abstract: Psychological attributes rarely operate in isolation: coaches and practitioners reason about networks of related traits rather than single indicators. We analyze a new dataset of 164 female volleyball players from Italy’s C and D leagues that combines standardized psychological profiling with background information. To learn directed relationships among mixed-type variables (ordinal questionnaire scores, categorical demographics, and continuous indicators), we introduce a hybrid structure learning approach that combines a latent Gaussian copula representation with a constraint-based skeleton and a score-based refinement to produce a single directed acyclic graph. We also study a bootstrap-aggregated variant to improve stability. In simulation studies spanning sample size, sparsity, and dimensionality, the proposed method achieves lower structural error and higher edge recovery than recent copula-based alternatives while maintaining high specificity.

Applied to volleyball, the learned network organizes mental skills around goal setting and self-confidence, with emotional arousal linking motivation and anxiety, and places key personality traits, most notably neuroticism and extraversion, upstream of skill clusters. Scenario analyses quantify how improvements in specific skills propagate through the network to shift preparation, confidence, and self-esteem. Overall, the approach provides an interpretable, data-driven framework for profiling psychological traits in sport and for supporting decisions in athlete development. This is joint work with Maria Iannario and Dae-Jin Lee.

Abstract: Hereditary melanoma is driven by germline pathogenic (PV) in high- and intermediate- to low penetrance susceptibility genes. Among these, a negative prognostic impact has been described for CDKN2A. However, the effect of other melanoma susceptibility genes on overall survival (OS) remains underexplored. Starting from a large international multicenter cohort of 2,253 patients, we investigated OS according to germline status of POT1/ACD, BAP1, MITF-E318K, pigmentation genes (TYR, OCA2, SLC45A2, TYRP1) and ATM. A total of 237 PV carriers were compared with matched non-carriers using Cox proportional hazards models adjusted for relevant confounders. Carriers of PV in POT1/ACD, BAP1, MITF-E318K, and ATM showed OS comparable to non carriers (HR range 0.50–3.65; all p>0.05). In contrast, carriers of PV in pigmentation genes (TYR, OCA2, SLC45A2, TYRP1) exhibited significantly improved OS (HR 0.27, 95% CI 0.08–0.94; p<0.05), with approximately halved all-cause mortality rates compared with non-carriers. Overall, germline PV in non-CDKN2A melanoma susceptibility genes were not associated with worse OS compared with non-carriers, whereas PV in pigmentation-related genes were associated with improved OS. These findings support the development of gene-specific prognostic models and tailored surveillance strategies. 

Abstract: 

Abstract: This work develops Bayesian switching models for analyzing high-dimensional time series with latent dynamic regimes. Across three distinct applications—dynamic brain connectivity in fMRI, human gesture phase segmentation from motion sensors, and circadian modeling of wearable and mobile data — we demonstrate how probabilistic state-switching frameworks uncover interpretable patterns in complex temporal systems. Together, these projects highlight the flexibility of switching models for capturing structured dynamics across neuroscience and behavioral data.

Abstract: Summer temperatures fluctuate dynamically around a seasonal component and may exhibit prolonged periods of extremely high temperatures, commonly referred to as heat waves. The classification of a period as a heat wave depends on the specific definition adopted, and it could be based on quantile-based thresholds. In this work, we address both heat wave detection and temperature dynamics modelling by leveraging a multi-state Bayesian Markov-switching regression model accounting for seasonality, long-term trends (via a large-scale climate index), and time-varying transition probabilities. By increasing the number of latent states, the model effectively isolates periods of high and extremely high temperatures, enabling the identification of heat wave episodes. We analyse the maximum daily temperatures across six geographically close but morphologically different weather stations of the Italian region of Friuli Venezia Giulia. We provide insights into regional variability in heat wave characteristics that reflect differences and similarities in the morphological conformation of the territory. 

Abstract: 

Abstract: 

Abstract: To summarise the adoption of healthy behaviours among the Italian adult population into a single, interpretable score.

• Using an institutional source (Data from the comprehensive Aspects of Daily Life (AVQ) multipurpose household survey conducted by the Italian National Statistical Institute (ISTAT)) 

• To be used for comparative purposes, both spatially and temporally

• Easily updatable

Abstract: Valos Srl è una CRO (Contract Research Organization) internazionale che opera principalmente in Europa e Nord America. Offre servizi focalizzati alla gestione del dato, con focus su Data Management, Programmazione statistica e Biostatistica. Le aree di maggior interesse degli studi clinici curati da Valos: Neurologia • Oncologia • Malattie infettive • Cardiovascolare • Epidemiologia • Respiratorio • Medical devices 

Abstract: Geographical analysis in epidemiological cancer research is an essential methodological tool for monitoring excess cancer occurrence in space and time and accumulating evidence useful to hypothesize the etiological impact of environmental factors (e.g., air pollution).

Abstract: