Development of unsupervised learning methods with applications to life sciences data

Machine Learning makes computers capable of performing tasks typically requiring human intelligence. A domain where it is having a considerable impact is the life sciences, allowing to devise new biological analysis protocols, develop patients’ treatments efficiently and faster, and reduce healthcare costs. This Thesis work presents new Machine Learning methods and pipelines for the life sciences focusing on the unsupervised field. At a methodological level, two methods are presented. The first is an “Ab Initio Local Principal Path” and it is a revised and improved version of a pre-existing algorithm in the manifold learning realm. The second contribution is an improvement over the Import Vector Domain Description (one-class learning) through the Kullback-Leibler divergence. It hybridizes kernel methods to Deep Learning obtaining a scalable solution, an improved probabilistic model, and state-of-the-art performances. Both methods are tested through several experiments, with a central focus on their relevance in life sciences. Results show that they improve the performances achieved by their previous versions. At the applicative level, two pipelines are presented. The first one is for the analysis of RNA-Seq datasets, both transcriptomic and single-cell data, and is aimed at identifying genes that may be involved in biological processes (e.g., the transition of tissues from normal to cancer). In this project, an R package is released on CRAN to make the pipeline accessible to the bioinformatic Community through high-level APIs. The second pipeline is in the drug discovery domain and is useful for identifying druggable pockets, namely regions of a protein with a high probability of accepting a small molecule (a drug). Both these pipelines achieve remarkable results. Lastly, a detour application is developed to identify the strengths/limitations of the “Principal Path” algorithm by analyzing Convolutional Neural Networks induced vector spaces. This application is conducted in the music and visual arts domains.

Design and implementation of bioinformatics tools for large scale genome annotation

The continuous increase of genome sequencing projects produced a huge amount of data in the last 10 years: currently more than 600 prokaryotic and 80 eukaryotic genomes are fully sequenced and publically available. However the sole sequencing process of a genome is able to determine just raw nucleotide sequences. This is only the first step of the genome annotation process that will deal with the issue of assigning biological information to each sequence. The annotation process is done at each different level of the biological information processing mechanism, from DNA to protein, and cannot be accomplished only by in vitro analysis procedures resulting extremely expensive and time consuming when applied at a this large scale level. Thus, in silico methods need to be used to accomplish the task. The aim of this work was the implementation of predictive computational methods to allow a fast, reliable, and automated annotation of genomes and proteins starting from aminoacidic sequences. The first part of the work was focused on the implementation of a new machine learning based method for the prediction of the subcellular localization of soluble eukaryotic proteins. The method is called BaCelLo, and was developed in 2006. The main peculiarity of the method is to be independent from biases present in the training dataset, which causes the over‐prediction of the most represented examples in all the other available predictors developed so far. This important result was achieved by a modification, made by myself, to the standard Support Vector Machine (SVM) algorithm with the creation of the so called Balanced SVM. BaCelLo is able to predict the most important subcellular localizations in eukaryotic cells and three, kingdom‐specific, predictors were implemented. In two extensive comparisons, carried out in 2006 and 2008, BaCelLo reported to outperform all the currently available state‐of‐the‐art methods for this prediction task. BaCelLo was subsequently used to completely annotate 5 eukaryotic genomes, by integrating it in a pipeline of predictors developed at the Bologna Biocomputing group by Dr. Pier Luigi Martelli and Dr. Piero Fariselli. An online database, called eSLDB, was developed by integrating, for each aminoacidic sequence extracted from the genome, the predicted subcellular localization merged with experimental and similarity‐based annotations. In the second part of the work a new, machine learning based, method was implemented for the prediction of GPI‐anchored proteins. Basically the method is able to efficiently predict from the raw aminoacidic sequence both the presence of the GPI‐anchor (by means of an SVM), and the position in the sequence of the post‐translational modification event, the so called ω‐site (by means of an Hidden Markov Model (HMM)). The method is called GPIPE and reported to greatly enhance the prediction performances of GPI‐anchored proteins over all the previously developed methods. GPIPE was able to predict up to 88% of the experimentally annotated GPI‐anchored proteins by maintaining a rate of false positive prediction as low as 0.1%. GPIPE was used to completely annotate 81 eukaryotic genomes, and more than 15000 putative GPI‐anchored proteins were predicted, 561 of which are found in H. sapiens. In average 1% of a proteome is predicted as GPI‐anchored. A statistical analysis was performed onto the composition of the regions surrounding the ω‐site that allowed the definition of specific aminoacidic abundances in the different considered regions. Furthermore the hypothesis that compositional biases are present among the four major eukaryotic kingdoms, proposed in literature, was tested and rejected. All the developed predictors and databases are freely available at: BaCelLo http://gpcr.biocomp.unibo.it/bacello eSLDB http://gpcr.biocomp.unibo.it/esldb GPIPE http://gpcr.biocomp.unibo.it/gpipe

Statistical methods for biomedical signal analysis and processing

Statistical modelling and statistical learning theory are two powerful analytical frameworks for analyzing signals and developing efficient processing and classification algorithms. In this thesis, these frameworks are applied for modelling and processing biomedical signals in two different contexts: ultrasound medical imaging systems and primate neural activity analysis and modelling. In the context of ultrasound medical imaging, two main applications are explored: deconvolution of signals measured from a ultrasonic transducer and automatic image segmentation and classification of prostate ultrasound scans. In the former application a stochastic model of the radio frequency signal measured from a ultrasonic transducer is derived. This model is then employed for developing in a statistical framework a regularized deconvolution procedure, for enhancing signal resolution. In the latter application, different statistical models are used to characterize images of prostate tissues, extracting different features. These features are then uses to segment the images in region of interests by means of an automatic procedure based on a statistical model of the extracted features. Finally, machine learning techniques are used for automatic classification of the different region of interests. In the context of neural activity signals, an example of bio-inspired dynamical network was developed to help in studies of motor-related processes in the brain of primate monkeys. The presented model aims to mimic the abstract functionality of a cell population in 7a parietal region of primate monkeys, during the execution of learned behavioural tasks.

Kernel Methods for Tree Structured Data

Machine learning comprises a series of techniques for automatic extraction of meaningful information from large collections of noisy data. In many real world applications, data is naturally represented in structured form. Since traditional methods in machine learning deal with vectorial information, they require an a priori form of preprocessing. Among all the learning techniques for dealing with structured data, kernel methods are recognized to have a strong theoretical background and to be effective approaches. They do not require an explicit vectorial representation of the data in terms of features, but rely on a measure of similarity between any pair of objects of a domain, the kernel function. Designing fast and good kernel functions is a challenging problem. In the case of tree structured data two issues become relevant: kernel for trees should not be sparse and should be fast to compute. The sparsity problem arises when, given a dataset and a kernel function, most structures of the dataset are completely dissimilar to one another. In those cases the classifier has too few information for making correct predictions on unseen data. In fact, it tends to produce a discriminating function behaving as the nearest neighbour rule. Sparsity is likely to arise for some standard tree kernel functions, such as the subtree and subset tree kernel, when they are applied to datasets with node labels belonging to a large domain. A second drawback of using tree kernels is the time complexity required both in learning and classification phases. Such a complexity can sometimes prevents the kernel application in scenarios involving large amount of data. This thesis proposes three contributions for resolving the above issues of kernel for trees. A first contribution aims at creating kernel functions which adapt to the statistical properties of the dataset, thus reducing its sparsity with respect to traditional tree kernel functions. Specifically, we propose to encode the input trees by an algorithm able to project the data onto a lower dimensional space with the property that similar structures are mapped similarly. By building kernel functions on the lower dimensional representation, we are able to perform inexact matchings between different inputs in the original space. A second contribution is the proposal of a novel kernel function based on the convolution kernel framework. Convolution kernel measures the similarity of two objects in terms of the similarities of their subparts. Most convolution kernels are based on counting the number of shared substructures, partially discarding information about their position in the original structure. The kernel function we propose is, instead, especially focused on this aspect. A third contribution is devoted at reducing the computational burden related to the calculation of a kernel function between a tree and a forest of trees, which is a typical operation in the classification phase and, for some algorithms, also in the learning phase. We propose a general methodology applicable to convolution kernels. Moreover, we show an instantiation of our technique when kernels such as the subtree and subset tree kernels are employed. In those cases, Direct Acyclic Graphs can be used to compactly represent shared substructures in different trees, thus reducing the computational burden and storage requirements.

Use of VEGFR-2 targeted microbubbles (BR55, Bracco imaging) for the early ultrasound evaluation of response to antiangiogenic treatment in a xenograft model of hepatocarcinoma

Aim: To evaluate the early response to treatment to an antiangiogenetic drug (sorafenib) in a heterotopic murine model of hepatocellular carcinoma (HCC) using ultrasonographic molecular imaging. Material and Methods: the xenographt model was established injecting a suspension of HuH7 cells subcutaneously in 19 nude mice. When tumors reached a mean diameter of 5-10 mm, they were divided in two groups (treatment and vehicle). The treatment group received sorafenib (62 mg/kg) by daily oral gavage for 14 days. Molecular imaging was performed using contrast enhanced ultrasound (CEUS), by injecting into the mouse venous circulation a suspension of VEGFR-2 targeted microbubbles (BR55, kind gift of Bracco Swiss, Geneve, Switzerland). Video clips were acquired for 6 minutes, then microbubbles (MBs) were destroyed by a high mechanical index (MI) impulse, and another minute was recorded to evaluate residual circulating MBs. The US protocol was repeated at day 0,+2,+4,+7, and +14 from the beginning of treatment administration. Video clips were analyzed using a dedicated software (Sonotumor, Bracco Swiss) to quantify the signal of the contrast agent. Time/intensity curves were obtained and the difference of the mean MBs signal before and after high MI impulse (Differential Targeted Enhancement-dTE) was calculated. dTE represents a numeric value in arbitrary units proportional to the amount of bound MBs. At day +14 mice were euthanized and the tumors analyzed for VEGFR-2, pERK, and CD31 tissue levels using western blot analysis. Results: dTE values decreased from day 0 to day +14 both in treatment and vehicle groups, and they were statistically higher in vehicle group than in treatment group at day +2, at day +7, and at day +14. With respect to the degree of tumor volume increase, measured as growth percentage delta (GPD), treatment group was divided in two sub-groups, non-responders (GPD>350%), and responders (GPD<200%). In the same way vehicle group was divided in slow growth group (GPD<400%), and fast growth group (GPD>900%). dTE values at day 0 (immediately before treatment start) were higher in non-responders than in responders group, with statistical difference at day 2. While dTE values were higher in the fast growth group than in the slow growth group only at day 0. A significant positive correlation was found between VEGFR-2 tissue levels and dTE values, confirming that level of BR55 tissue enhancement reflects the amount of tissue VEGF receptor. Conclusions: the present findings show that, at least in murine experimental models, CEUS with BR55 is feasable and appears to be a useful tool in the prediction of tumor growth and response to sorafenib treatment in xenograft HCC.

Cambiamenti dello scenario epidemiologico e clinico dell'epatocarcinoma in Italia nell'ultimo ventennio

Background & Aims: This study investigates whether the aetiologic changes in liver disease and the improved management of hepatocellular carcinoma (HCC) have modified the clinical scenario of this tumour over the last 20 years in Italy. Methods: Retrospective study based on the analysis of the ITA.LI.CA (Italian Liver Cancer) database including 3027 HCC patients managed in 11 centres. Patients were divided into 3 groups according to the period of HCC diagnosis: 1987–1996 (year of the ‘‘Milano criteria’’ publication), 1997–2001 (year of release of the EASL guidelines for HCC), and 2002–2008. Results: The significant changes were: (1) progressive patient ageing; (2) increasing prevalence of HCV infection until 2001, with a subsequent decrease, when the alcoholic aetiology increased; (3) liver function improvement, until 2001; (4) increasing ‘‘incidental’’ at the expense of ‘‘symptomatic’’ diagnoses, until 2001; (5) unchanged prevalence of tumours diagnosed during surveillance (around 50%), with an increasing use of the 6- month schedule; (6) favourable HCC ‘‘stage migration’’, until 2001; (7) increasing use of percutaneous ablation; (8) improving survival, until 2001. Conclusions: Over the last 20 years, several aetiologic and clinical features regarding HCC have changed. The survival improvement observed until 2001 was due to an increasing number of tumours diagnosed in early stages and in a background of compensated cirrhosis, and a growing and better use of locoregional treatments. However, the prevalence of early cancers and survival did not increase further in the last years, a result inciting national policies aimed at implementing surveillance programmes for at risk patients.

Valutazione precoce della risposta a Sorafenib nel Carcinoma Epatocellulare mediante quantificazione dell'Ecografia con contrasto con software Sonotumor

Scopo dello studio: Stabilire se cambiamenti della perfusione di una lesione target di epatocarcinoma (HCC), valutati quantitativamente mediante ecografia con contrasto (CE-US) alla settimana 2 e 4 di terapia con sorafenib, possono predire la progressione di malattia alla settimana 8, valutata con la tomografia computerizzata o la risonanza magnetica con mezzo di contrasto (TC-RM) usando i criteri RECIST/RECIST modificati (response evaluation criteria in solid tumors). Pazienti e metodi: Il comitato etico ha approvato lo studio ed i pazienti hanno fornito un consenso informato scritto prima dell’arruolamento. Lo studio è stato effettuato su un campione di soggetti con epatocarcinoma avanzato o non suscettibile di trattamento curativo, in monoterapia con sorafenib. La valutazione della risposta tumorale è stata effettuata con TC o RM a 2 mesi usando i criteri RECIST/RECIST modificati. La CE-US è stata effettuata entro 1 settimana prima dell’inizio del trattamento con sorafenib e durante la terapia alla settimana 2, 4, 8, 16 e 32. I parametri quantitativi funzionali sono stati ottenuti impiegando un software dedicato. I cambiamenti dei valori dei parametri suddetti tra il tempo zero ed i punti temporali successivi sono stati confrontati con la risposta tumorale basata sui criteri RECIST/RECIST modificati. Risultati: La riduzione dei valori dei parametri relativi alla perfusione tumorale, in particolare di WiAUC e PE (parametri correlati con il volume ematico), al T2/T4 (settimana 2, 4), predice la risposta tumorale a 2 mesi, valutata secondo i criteri RECIST e RECIST modificati, risultata indicativa di malattia stabile (responders). Conclusione: L’ecografia con contrasto può essere impiegata per quantificare i cambiamenti della vascolarizzazione tumorale già alla settimana 2, 4 dopo la somministrazione di sorafenib nei pazienti con HCC. Questi precoci cambiamenti della perfusione tumorale possono essere predittivi della risposta tumorale a 2 mesi e possono avere un potenziale nella valutazione precoce dell'efficacia della terapia antiangiogenica nell’epatocarcinoma.

Risposta Immunitaria Epatite B specifica in pazienti affetti da Epatocarcinoma

Hepatitis B x protein (HBx) is a non structural, multifunctional protein of hepatitis B virus (HBV) that modulates a variety of host processes.Due to its transcriptional activity,able to alter the expression of growth-control genes,it has been implicated in hepatocarcinogenesis.Increased expression of HBx has been reported on the liver tissue samples of hepatocellular carcinoma (HCC),and a specific anti-HBx immune response can be detected in the peripheral blood of patients with chronic HBV.However,its role and entity has not been yet clarified.Thus,we performed a cross-sectional analysis of anti-HBx specific T cell response in HBV-infected patients in different stage of disease.A total of 70 HBV-infected subjects were evaluated:15 affected by chronic hepatitis (CH-median age 45 yrs),14 by cirrhosis (median age 55 yrs),11 with dysplastic nodules (median age 64 yrs),15 with HCC (median age 60 yrs),15 with IC(median age 53 yrs).All patients were infected by virus genotype D with different levels of HBV viremia and most of them (91%) were HBeAb positive.The HBx-specific T cell response was evaluated by anti-Interferon (IFN)-gamma Elispot assay after in vitro stimulation of peripheral blood mononuclear cells,using 20 overlapping synthetic peptides covering all HBx protein sequence.HBx-specific IFN-gamma-secreting T cells were found in 6 out of 15 patients with chronic hepatitis (40%), 3 out of 14 cirrhosis (21%), in 5 out of 11 cirrhosis with macronodules (54%), and in 10 out of 15 HCC patients (67%). The number of responding patients resulted significantly higher in HCC than IC (p=0.02) and cirrhosis (p=0.02). Central specific region of the protein x was preferentially recognize,between 86-88 peptides. HBx response does not correlate with clinical feature disease(AFP,MELD).The HBx specific T-cell response seems to increase accordingly to progression of the disease, being increased in subjects with dysplastic or neoplastic lesions and can represent an additional tool to monitor the patients at high risk to develop HCC

Aspetti diagnostici, strategie terapeutiche ed esito clinico dei tumori primitivi colangiocellulari insorti su cirrosi

Background: The recent increasing incidence of intrahepatic cholangiocellular carcinoma (ICC) in cirrhosis increased the problem of noninvasive differential diagnosis between ICC and hepatocellular carcinoma (HCC) in cirrhosis. In literature there isn’t data about treatment and prognosis of ICC in cirrhosis. Aim: To investigate the role of the different imaging techniques in the diagnosis of ICC in cirrhosis; to analyze treatments and prognosis with particular attention to factors associated with survival. Methods: The data of 30 cirrhotic patients with ICC were retrospectively collected; patients were referred to Liver Units (S.Orsola-Malpighi and S.Matteo Hospitals) between 2005 and 2011. The results of contrast-enhanced ultrasound (CEUS), computed tomography (CT) and magnetic resonance (MR) were evaluated; the enhancement pattern at different imaging techniques were analysed, with particular attention to misdiagnosis of HCC. We evaluated the different treatments and survival of the study group and then we performed the survival analysis of different clinico-pathologic factors. Results: Twenty-five patients underwent CEUS, 27 CT and 10 MR. In 3 cases (12%) CEUS misdiagnosed ICC for HCC, in 7 cases (26%) CT misdiagnosed ICC and in 1 case (10%) MR misdiagnosed ICC. Patient were followed for a mean of 30 months (range:4-86), with a mean survival of 30 months. Twenty-four out of 30 patients were treated with curative approach, while the other 6 underwent TACE (n=4), radioembolization (n=1) or systemic treatment with Gemcitabine (n=1). The univariate analysis revealed that CA19-9 levels, surveillance program and nodule size were significantly related with survival. By multivariate analysis only nodule size £ 40mm was significant (p=0,004). Conclusion: Diagnosis of ICC in cirrhosis remains difficult because there isn’t a typical enhancement pattern and in some cases it cannot be distinguished from HCC by the different imaging techniques. The study of survival related factors shows that nodule size ≤ 40mm is correlated with improved survival.

Lesion detection and classification in breast cancer: evaluation of approaches based on morphological features, tracer kinetic modelling and semi-quantitative parameters in MR functional imaging (DCE-MRI)

The diagnosis, grading and classification of tumours has benefited considerably from the development of DCE-MRI which is now essential to the adequate clinical management of many tumour types due to its capability in detecting active angiogenesis. Several strategies have been proposed for DCE-MRI evaluation. Visual inspection of contrast agent concentration curves vs time is a very simple yet operator dependent procedure, therefore more objective approaches have been developed in order to facilitate comparison between studies. In so called model free approaches, descriptive or heuristic information extracted from time series raw data have been used for tissue classification. The main issue concerning these schemes is that they have not a direct interpretation in terms of physiological properties of the tissues. On the other hand, model based investigations typically involve compartmental tracer kinetic modelling and pixel-by-pixel estimation of kinetic parameters via non-linear regression applied on region of interests opportunely selected by the physician. This approach has the advantage to provide parameters directly related to the pathophysiological properties of the tissue such as vessel permeability, local regional blood flow, extraction fraction, concentration gradient between plasma and extravascular-extracellular space. Anyway, nonlinear modelling is computational demanding and the accuracy of the estimates can be affected by the signal-to-noise ratio and by the initial solutions. The principal aim of this thesis is investigate the use of semi-quantitative and quantitative parameters for segmentation and classification of breast lesion. The objectives can be subdivided as follow: describe the principal techniques to evaluate time intensity curve in DCE-MRI with focus on kinetic model proposed in literature; to evaluate the influence in parametrization choice for a classic bi-compartmental kinetic models; to evaluate the performance of a method for simultaneous tracer kinetic modelling and pixel classification; to evaluate performance of machine learning techniques training for segmentation and classification of breast lesion.

Automatic risk evaluation in elderly patients based on Autonomic Nervous System assessment

Dysfunction of Autonomic Nervous System (ANS) is a typical feature of chronic heart failure and other cardiovascular disease. As a simple non-invasive technology, heart rate variability (HRV) analysis provides reliable information on autonomic modulation of heart rate. The aim of this thesis was to research and develop automatic methods based on ANS assessment for evaluation of risk in cardiac patients. Several features selection and machine learning algorithms have been combined to achieve the goals. Automatic assessment of disease severity in Congestive Heart Failure (CHF) patients: a completely automatic method, based on long-term HRV was proposed in order to automatically assess the severity of CHF, achieving a sensitivity rate of 93% and a specificity rate of 64% in discriminating severe versus mild patients. Automatic identification of hypertensive patients at high risk of vascular events: a completely automatic system was proposed in order to identify hypertensive patients at higher risk to develop vascular events in the 12 months following the electrocardiographic recordings, achieving a sensitivity rate of 71% and a specificity rate of 86% in identifying high-risk subjects among hypertensive patients. Automatic identification of hypertensive patients with history of fall: it was explored whether an automatic identification of fallers among hypertensive patients based on HRV was feasible. The results obtained in this thesis could have implications both in clinical practice and in clinical research. The system has been designed and developed in order to be clinically feasible. Moreover, since 5-minute ECG recording is inexpensive, easy to assess, and non-invasive, future research will focus on the clinical applicability of the system as a screening tool in non-specialized ambulatories, in order to identify high-risk patients to be shortlisted for more complex investigations.

Prospettive di terapia sistemica combinata sorafenib + capecitabina per l'epatocarcinoma. Valutazione nel modello animale e ruolo dell'imaging ad ultrasuoni

Il Sorafenib è l’unica terapia sistemica approvata per l’epatocarcinoma (HCC) avanzato. Tuttavia, molti tumori sviluppano resistenze. La chemioterapia metronomica sembrerebbe avere un effetto antiangiogenetico. La Capecitabina metronomica è potenzialmente efficace nell’HCC avanzato. Lo scopo dello studio è stato valutare il comportamento di un modello murino di HCC sottoposto a Sorafenib, Capecitabina e terapia combinata, per dimostrarne un eventuale effetto sinergico. Il modello è stato creato in topi scid mediante inoculazione sottocutanea di 5 milioni di cellule HuH7. I topi sono stati suddivisi in 4 gruppi: gruppo 1 sottoposto a terapia con placebo (9 topi), gruppo 2 a Sorafenib (7 topi), gruppo 3 a Capecitabina (7 topi) e gruppo 4 a terapia combinata Sorafenib+Capecitabina (10 topi). I topi sono stati studiati al giorno 0 e 14 con ecografia B-mode e con mezzo di contrasto (CEUS). Al giorno 14 sono stati sacrificati e i pezzi tumorali sono stati conservati per l’analisi Western Blot. Un topo del gruppo 1 e 4 topi del gruppo 4 sono morti precocemente e quindi sono stati esclusi. Il delta di crescita tumorale al giorno 14 rispetto al giorno 0 è risultato di +503 %, +158 %, +462 % e +176 % rispettivamente nei 4 gruppi (p<0.05 tra i 4 gruppi, tra il gruppo 1 e 2, tra il gruppo 1 e 4, tra il gruppo 2 e 3, tra il gruppo 3 e 4). Alla CEUS non si sono evidenziate differenze statisticamente significative nei cambiamenti di perfusione tumorale al giorno 14 nei 4 gruppi. L’analisi Western Blot ha mostrato livelli di VEGFR-2 inferiori nel gruppo dei topi trattati con Sorafenib. La terapia di associazione di Sorafenib e Capecitabina non comporta un beneficio, in termini di riduzione della crescita tumorale, in un modello murino di HCC rispetto al solo Sorafenib. Inoltre, può essere sospettato un incremento di tossicità.

Cambiamenti epidemiologici dell'epatocarcinoma agli inizi del xxi secolo: Dal ruolo emergente delle malattie dismetaboliche quali fattori eziopatogemetici alla necessita' di revisione del sistema di stadiaziome

Introduzione e scopo: la rapida diffusione delle malattie dismetaboliche sta modificando l’epidemiologia dell’epatocarcinoma (HCC). Scopo della tesi è, attraverso quattro studi, analizzare l’impatto di questi cambiamenti nella gestione clinica del paziente affetto da HCC. Materiali e metodi: quattro studi di coorte, condotti con analisi retrospettiva del database ITA.LI.CA. Studio 1:3658 pazienti arruolati tra il 01-01-2001 ed il 31-12-2012 suddivisi in base alla data di diagnosi:2001-2004 (954 pazienti), 2005-2008 (1122 pazienti), 2009-2012 (1582 pazienti). Studio 2:analisi comparativa tra 756 pazienti con HCC-NAFLD e 611 pazienti con HCC-HCV. Studio 3:proposta di quattro modelli alternativi al BCLC originale con validazione di una proposta di sottostadiazione dell’intermedio, considerando 2606 pazienti arruolati tra il 01-01-2000 e il 31-12-2012 e riallocati secondo gradi diversi di perfomance status (PS). Studio 4:analisi di 696 pazienti con HCC in stadio intermedio diagnosticato dopo il 1999 stratificati per trattamento. Risultati: studio 1:progressivo aumento dell’età alla diagnosi e delle eziologie dismetaboliche; più frequente esordio dell’HCC in stadio precoce e con funzione epatica più conservata; aumento della sopravvivenza dopo il 2008. Studio 2:i pazienti con HCC-NAFLD mostrano più frequentemente un tumore infiltrativo diagnosticato fuori dai programmi di sorveglianza, con prognosi peggiore rispetto ai pazienti HCC-HCV. Questa differenza di sopravvivenza si elimina rimuovendo i fattori di confondimento attraverso propensity analysis. Studio 3:il PS1 non è un predittore indipendente di sopravvivenza. Il modello 4 (considerando PS1=PS0 e con la sottostadiazione proposta), ha la migliore capacità discriminativa. Studio 4:i trattamenti curativi riducono la mortalità più della TACE, anche dopo propensity analysis. Conclusioni: l’aumento delle patologie dismetaboliche comporterà diagnosi di malattia ad uno stadio più avanzato, quando sintomatica, rendendo necessario stabilire un programma di sorveglianza. Inoltre per una migliore stratificazione e gestione dei pazienti, bisogna riconsiderare il ruolo del PS ed offrire un ventaglio di opzioni terapeutiche anche per il pazienti in stadio intermedio.