A communication infrastructure to support knowledge level agents on the web

Agent Communication Languages (ACLs) have been developed to provide a way for agents to communicate with each other supporting cooperation in Multi-Agent Systems. In the past few years many ACLs have been proposed for Multi-Agent Systems, such as KQML and FIPA-ACL. The goal of these languages is to support high-level, human like communication among agents, exploiting Knowledge Level features rather than symbol level ones. Adopting these ACLs, and mainly the FIPA-ACL specifications, many agent platforms and prototypes have been developed. Despite these efforts, an important issue in the research on ACLs is still open and concerns how these languages should deal (at the Knowledge Level) with possible failures of agents. Indeed, the notion of Knowledge Level cannot be straightforwardly extended to a distributed framework such as MASs, because problems concerning communication and concurrency may arise when several Knowledge Level agents interact (for example deadlock or starvation). The main contribution of this Thesis is the design and the implementation of NOWHERE, a platform to support Knowledge Level Agents on the Web. NOWHERE exploits an advanced Agent Communication Language, FT-ACL, which provides high-level fault-tolerant communication primitives and satisfies a set of well defined Knowledge Level programming requirements. NOWHERE is well integrated with current technologies, for example providing full integration for Web services. Supporting different middleware used to send messages, it can be adapted to various scenarios. In this Thesis we present the design and the implementation of the architecture, together with a discussion of the most interesting details and a comparison with other emerging agent platforms. We also present several case studies where we discuss the benefits of programming agents using the NOWHERE architecture, comparing the results with other solutions. Finally, the complete source code of the basic examples can be found in appendix.

New approaches to open problems in gene expression microarray data

In the past decade, the advent of efficient genome sequencing tools and high-throughput experimental biotechnology has lead to enormous progress in the life science. Among the most important innovations is the microarray tecnology. It allows to quantify the expression for thousands of genes simultaneously by measurin the hybridization from a tissue of interest to probes on a small glass or plastic slide. The characteristics of these data include a fair amount of random noise, a predictor dimension in the thousand, and a sample noise in the dozens. One of the most exciting areas to which microarray technology has been applied is the challenge of deciphering complex disease such as cancer. In these studies, samples are taken from two or more groups of individuals with heterogeneous phenotypes, pathologies, or clinical outcomes. these samples are hybridized to microarrays in an effort to find a small number of genes which are strongly correlated with the group of individuals. Eventhough today methods to analyse the data are welle developed and close to reach a standard organization (through the effort of preposed International project like Microarray Gene Expression Data -MGED- Society [1]) it is not unfrequant to stumble in a clinician's question that do not have a compelling statistical method that could permit to answer it.The contribution of this dissertation in deciphering disease regards the development of new approaches aiming at handle open problems posed by clinicians in handle specific experimental designs. In Chapter 1 starting from a biological necessary introduction, we revise the microarray tecnologies and all the important steps that involve an experiment from the production of the array, to the quality controls ending with preprocessing steps that will be used into the data analysis in the rest of the dissertation. While in Chapter 2 a critical review of standard analysis methods are provided stressing most of problems that In Chapter 3 is introduced a method to adress the issue of unbalanced design of miacroarray experiments. In microarray experiments, experimental design is a crucial starting-point for obtaining reasonable results. In a two-class problem, an equal or similar number of samples it should be collected between the two classes. However in some cases, e.g. rare pathologies, the approach to be taken is less evident. We propose to address this issue by applying a modified version of SAM [2]. MultiSAM consists in a reiterated application of a SAM analysis, comparing the less populated class (LPC) with 1,000 random samplings of the same size from the more populated class (MPC) A list of the differentially expressed genes is generated for each SAM application. After 1,000 reiterations, each single probe given a "score" ranging from 0 to 1,000 based on its recurrence in the 1,000 lists as differentially expressed. The performance of MultiSAM was compared to the performance of SAM and LIMMA [3] over two simulated data sets via beta and exponential distribution. The results of all three algorithms over low- noise data sets seems acceptable However, on a real unbalanced two-channel data set reagardin Chronic Lymphocitic Leukemia, LIMMA finds no significant probe, SAM finds 23 significantly changed probes but cannot separate the two classes, while MultiSAM finds 122 probes with score >300 and separates the data into two clusters by hierarchical clustering. We also report extra-assay validation in terms of differentially expressed genes Although standard algorithms perform well over low-noise simulated data sets, multi-SAM seems to be the only one able to reveal subtle differences in gene expression profiles on real unbalanced data. In Chapter 4 a method to adress similarities evaluation in a three-class prblem by means of Relevance Vector Machine [4] is described. In fact, looking at microarray data in a prognostic and diagnostic clinical framework, not only differences could have a crucial role. In some cases similarities can give useful and, sometimes even more, important information. The goal, given three classes, could be to establish, with a certain level of confidence, if the third one is similar to the first or the second one. In this work we show that Relevance Vector Machine (RVM) [2] could be a possible solutions to the limitation of standard supervised classification. In fact, RVM offers many advantages compared, for example, with his well-known precursor (Support Vector Machine - SVM [3]). Among these advantages, the estimate of posterior probability of class membership represents a key feature to address the similarity issue. This is a highly important, but often overlooked, option of any practical pattern recognition system. We focused on Tumor-Grade-three-class problem, so we have 67 samples of grade I (G1), 54 samples of grade 3 (G3) and 100 samples of grade 2 (G2). The goal is to find a model able to separate G1 from G3, then evaluate the third class G2 as test-set to obtain the probability for samples of G2 to be member of class G1 or class G3. The analysis showed that breast cancer samples of grade II have a molecular profile more similar to breast cancer samples of grade I. Looking at the literature this result have been guessed, but no measure of significance was gived before.

Caratterizzazione del legame di molecole di interesse farmaceutico alla sieroalbumina umana mediante biocromatografia e dicroismo circolare

Negli ultimi anni, un crescente numero di studiosi ha focalizzato la propria attenzione sullo sviluppo di strategie che permettessero di caratterizzare le proprietà ADMET dei farmaci in via di sviluppo, il più rapidamente possibile. Questa tendenza origina dalla consapevolezza che circa la metà dei farmaci in via di sviluppo non viene commercializzato perché ha carenze nelle caratteristiche ADME, e che almeno la metà delle molecole che riescono ad essere commercializzate, hanno comunque qualche problema tossicologico o ADME [1]. Infatti, poco importa quanto una molecola possa essere attiva o specifica: perché possa diventare farmaco è necessario che venga ben assorbita, distribuita nell’organismo, metabolizzata non troppo rapidamente, ne troppo lentamente e completamente eliminata. Inoltre la molecola e i suoi metaboliti non dovrebbero essere tossici per l’organismo. Quindi è chiaro come una rapida determinazione dei parametri ADMET in fasi precoci dello sviluppo del farmaco, consenta di risparmiare tempo e denaro, permettendo di selezionare da subito i composti più promettenti e di lasciar perdere quelli con caratteristiche negative. Questa tesi si colloca in questo contesto, e mostra l’applicazione di una tecnica semplice, la biocromatografia, per caratterizzare rapidamente il legame di librerie di composti alla sieroalbumina umana (HSA). Inoltre mostra l’utilizzo di un’altra tecnica indipendente, il dicroismo circolare, che permette di studiare gli stessi sistemi farmaco-proteina, in soluzione, dando informazioni supplementari riguardo alla stereochimica del processo di legame. La HSA è la proteina più abbondante presente nel sangue. Questa proteina funziona da carrier per un gran numero di molecole, sia endogene, come ad esempio bilirubina, tiroxina, ormoni steroidei, acidi grassi, che xenobiotici. Inoltre aumenta la solubilità di molecole lipofile poco solubili in ambiente acquoso, come ad esempio i tassani. Il legame alla HSA è generalmente stereoselettivo e ad avviene a livello di siti di legame ad alta affinità. Inoltre è ben noto che la competizione tra farmaci o tra un farmaco e metaboliti endogeni, possa variare in maniera significativa la loro frazione libera, modificandone l’attività e la tossicità. Per queste sue proprietà la HSA può influenzare sia le proprietà farmacocinetiche che farmacodinamiche dei farmaci. Non è inusuale che un intero progetto di sviluppo di un farmaco possa venire abbandonato a causa di un’affinità troppo elevata alla HSA, o a un tempo di emivita troppo corto, o a una scarsa distribuzione dovuta ad un debole legame alla HSA. Dal punto di vista farmacocinetico, quindi, la HSA è la proteina di trasporto del plasma più importante. Un gran numero di pubblicazioni dimostra l’affidabilità della tecnica biocromatografica nello studio dei fenomeni di bioriconoscimento tra proteine e piccole molecole [2-6]. Il mio lavoro si è focalizzato principalmente sull’uso della biocromatografia come metodo per valutare le caratteristiche di legame di alcune serie di composti di interesse farmaceutico alla HSA, e sul miglioramento di tale tecnica. Per ottenere una miglior comprensione dei meccanismi di legame delle molecole studiate, gli stessi sistemi farmaco-HSA sono stati studiati anche con il dicroismo circolare (CD). Inizialmente, la HSA è stata immobilizzata su una colonna di silice epossidica impaccata 50 x 4.6 mm di diametro interno, utilizzando una procedura precedentemente riportata in letteratura [7], con alcune piccole modifiche. In breve, l’immobilizzazione è stata effettuata ponendo a ricircolo, attraverso una colonna precedentemente impaccata, una soluzione di HSA in determinate condizioni di pH e forza ionica. La colonna è stata quindi caratterizzata per quanto riguarda la quantità di proteina correttamente immobilizzata, attraverso l’analisi frontale di L-triptofano [8]. Di seguito, sono stati iniettati in colonna alcune soluzioni raceme di molecole note legare la HSA in maniera enantioselettiva, per controllare che la procedura di immobilizzazione non avesse modificato le proprietà di legame della proteina. Dopo essere stata caratterizzata, la colonna è stata utilizzata per determinare la percentuale di legame di una piccola serie di inibitori della proteasi HIV (IPs), e per individuarne il sito(i) di legame. La percentuale di legame è stata calcolata attraverso il fattore di capacità (k) dei campioni. Questo parametro in fase acquosa è stato estrapolato linearmente dal grafico log k contro la percentuale (v/v) di 1-propanolo presente nella fase mobile. Solamente per due dei cinque composti analizzati è stato possibile misurare direttamente il valore di k in assenza di solvente organico. Tutti gli IPs analizzati hanno mostrato un’elevata percentuale di legame alla HSA: in particolare, il valore per ritonavir, lopinavir e saquinavir è risultato maggiore del 95%. Questi risultati sono in accordo con dati presenti in letteratura, ottenuti attraverso il biosensore ottico [9]. Inoltre, questi risultati sono coerenti con la significativa riduzione di attività inibitoria di questi composti osservata in presenza di HSA. Questa riduzione sembra essere maggiore per i composti che legano maggiormente la proteina [10]. Successivamente sono stati eseguiti degli studi di competizione tramite cromatografia zonale. Questo metodo prevede di utilizzare una soluzione a concentrazione nota di un competitore come fase mobile, mentre piccole quantità di analita vengono iniettate nella colonna funzionalizzata con HSA. I competitori sono stati selezionati in base al loro legame selettivo ad uno dei principali siti di legame sulla proteina. In particolare, sono stati utilizzati salicilato di sodio, ibuprofene e valproato di sodio come marker dei siti I, II e sito della bilirubina, rispettivamente. Questi studi hanno mostrato un legame indipendente dei PIs ai siti I e II, mentre è stata osservata una debole anticooperatività per il sito della bilirubina. Lo stesso sistema farmaco-proteina è stato infine investigato in soluzione attraverso l’uso del dicroismo circolare. In particolare, è stato monitorata la variazione del segnale CD indotto di un complesso equimolare [HSA]/[bilirubina], a seguito dell’aggiunta di aliquote di ritonavir, scelto come rappresentante della serie. I risultati confermano la lieve anticooperatività per il sito della bilirubina osservato precedentemente negli studi biocromatografici. Successivamente, lo stesso protocollo descritto precedentemente è stato applicato a una colonna di silice epossidica monolitica 50 x 4.6 mm, per valutare l’affidabilità del supporto monolitico per applicazioni biocromatografiche. Il supporto monolitico monolitico ha mostrato buone caratteristiche cromatografiche in termini di contropressione, efficienza e stabilità, oltre che affidabilità nella determinazione dei parametri di legame alla HSA. Questa colonna è stata utilizzata per la determinazione della percentuale di legame alla HSA di una serie di poliamminochinoni sviluppati nell’ambito di una ricerca sulla malattia di Alzheimer. Tutti i composti hanno mostrato una percentuale di legame superiore al 95%. Inoltre, è stata osservata una correlazione tra percentuale di legame è caratteristiche della catena laterale (lunghezza e numero di gruppi amminici). Successivamente sono stati effettuati studi di competizione dei composti in esame tramite il dicroismo circolare in cui è stato evidenziato un effetto anticooperativo dei poliamminochinoni ai siti I e II, mentre rispetto al sito della bilirubina il legame si è dimostrato indipendente. Le conoscenze acquisite con il supporto monolitico precedentemente descritto, sono state applicate a una colonna di silice epossidica più corta (10 x 4.6 mm). Il metodo di determinazione della percentuale di legame utilizzato negli studi precedenti si basa su dati ottenuti con più esperimenti, quindi è necessario molto tempo prima di ottenere il dato finale. L’uso di una colonna più corta permette di ridurre i tempi di ritenzione degli analiti, per cui la determinazione della percentuale di legame alla HSA diventa molto più rapida. Si passa quindi da una analisi a medio rendimento a una analisi di screening ad alto rendimento (highthroughput- screening, HTS). Inoltre, la riduzione dei tempi di analisi, permette di evitare l’uso di soventi organici nella fase mobile. Dopo aver caratterizzato la colonna da 10 mm con lo stesso metodo precedentemente descritto per le altre colonne, sono stati iniettati una serie di standard variando il flusso della fase mobile, per valutare la possibilità di utilizzare flussi elevati. La colonna è stata quindi impiegata per stimare la percentuale di legame di una serie di molecole con differenti caratteristiche chimiche. Successivamente è stata valutata la possibilità di utilizzare una colonna così corta, anche per studi di competizione, ed è stata indagato il legame di una serie di composti al sito I. Infine è stata effettuata una valutazione della stabilità della colonna in seguito ad un uso estensivo. L’uso di supporti cromatografici funzionalizzati con albumine di diversa origine (ratto, cane, guinea pig, hamster, topo, coniglio), può essere proposto come applicazione futura di queste colonne HTS. Infatti, la possibilità di ottenere informazioni del legame dei farmaci in via di sviluppo alle diverse albumine, permetterebbe un migliore paragone tra i dati ottenuti tramite esperimenti in vitro e i dati ottenuti con esperimenti sull’animale, facilitando la successiva estrapolazione all’uomo, con la velocità di un metodo HTS. Inoltre, verrebbe ridotto anche il numero di animali utilizzati nelle sperimentazioni. Alcuni lavori presenti in letteratura dimostrano l’affidabilita di colonne funzionalizzate con albumine di diversa origine [11-13]: l’utilizzo di colonne più corte potrebbe aumentarne le applicazioni.

Anticancer drug screening from images of zebrafish embryogenesis

During the previous 10 years, global R&D expenditure in the pharmaceuticals and biotechnology sector has steadily increased, without a corresponding increase in output of new medicines. To address this situation, the biopharmaceutical industry's greatest need is to predict the failures at the earliest possible stage of the drug development process. A major key to reducing failures in drug screenings is the development and use of preclinical models that are more predictive of efficacy and safety in clinical trials. Further, relevant animal models are needed to allow a wider testing of novel hypotheses. Key to this is the developing, refining, and validating of complex animal models that directly link therapeutic targets to the phenotype of disease, allowing earlier prediction of human response to medicines and identification of safety biomarkers. Morehover, well-designed animal studies are essential to bridge the gap between test in cell cultures and people. Zebrafish is emerging, complementary to other models, as a powerful system for cancer studies and drugs discovery. We aim to investigate this research area designing a new preclinical cancer model based on the in vivo imaging of zebrafish embryogenesis. Technological advances in imaging have made it feasible to acquire nondestructive in vivo images of fluorescently labeled structures, such as cell nuclei and membranes, throughout early Zebrafishsh embryogenesis. This In vivo image-based investigation provides measurements for a large number of features at cellular level and events including nuclei movements, cells counting, and mitosis detection, thereby enabling the estimation of more significant parameters such as proliferation rate, highly relevant for investigating anticancer drug effects. In this work, we designed a standardized procedure for accessing drug activity at the cellular level in live zebrafish embryos. The procedure includes methodologies and tools that combine imaging and fully automated measurements of embryonic cell proliferation rate. We achieved proliferation rate estimation through the automatic classification and density measurement of epithelial enveloping layer and deep layer cells. Automatic embryonic cells classification provides the bases to measure the variability of relevant parameters, such as cell density, in different classes of cells and is finalized to the estimation of efficacy and selectivity of anticancer drugs. Through these methodologies we were able to evaluate and to measure in vivo the therapeutic potential and overall toxicity of Dbait and Irinotecan anticancer molecules. Results achieved on these anticancer molecules are presented and discussed; furthermore, extensive accuracy measurements are provided to investigate the robustness of the proposed procedure. Altogether, these observations indicate that zebrafish embryo can be a useful and cost-effective alternative to some mammalian models for the preclinical test of anticancer drugs and it might also provides, in the near future, opportunities to accelerate the process of drug discovery.

Development of original analytical methods for the therapeutic drug monitoring of CNS druges: Antipsychotics, Antidepressants and Anxiolytics-hypnotics

Great strides have been made in the last few years in the pharmacological treatment of neuropsychiatric disorders, with the introduction into the therapy of several new and more efficient agents, which have improved the quality of life of many patients. Despite these advances, a large percentage of patients is still considered “non-responder” to the therapy, not drawing any benefits from it. Moreover, these patients have a peculiar therapeutic profile, due to the very frequent application of polypharmacy, attempting to obtain satisfactory remission of the multiple aspects of psychiatric syndromes. Therapy is heavily individualised and switching from one therapeutic agent to another is quite frequent. One of the main problems of this situation is the possibility of unwanted or unexpected pharmacological interactions, which can occur both during polypharmacy and during switching. Simultaneous administration of psychiatric drugs can easily lead to interactions if one of the administered compounds influences the metabolism of the others. Impaired CYP450 function due to inhibition of the enzyme is frequent. Other metabolic pathways, such as glucuronidation, can also be influenced. The Therapeutic Drug Monitoring (TDM) of psychotropic drugs is an important tool for treatment personalisation and optimisation. It deals with the determination of parent drugs and metabolites plasma levels, in order to monitor them over time and to compare these findings with clinical data. This allows establishing chemical-clinical correlations (such as those between administered dose and therapeutic and side effects), which are essential to obtain the maximum therapeutic efficacy, while minimising side and toxic effects. It is evident the importance of developing sensitive and selective analytical methods for the determination of the administered drugs and their main metabolites, in order to obtain reliable data that can correctly support clinical decisions. During the three years of Ph.D. program, some analytical methods based on HPLC have been developed, validated and successfully applied to the TDM of psychiatric patients undergoing treatment with drugs belonging to following classes: antipsychotics, antidepressants and anxiolytic-hypnotics. The biological matrices which have been processed were: blood, plasma, serum, saliva, urine, hair and rat brain. Among antipsychotics, both atypical and classical agents have been considered, such as haloperidol, chlorpromazine, clotiapine, loxapine, risperidone (and 9-hydroxyrisperidone), clozapine (as well as N-desmethylclozapine and clozapine N-oxide) and quetiapine. While the need for an accurate TDM of schizophrenic patients is being increasingly recognized by psychiatrists, only in the last few years the same attention is being paid to the TDM of depressed patients. This is leading to the acknowledgment that depression pharmacotherapy can greatly benefit from the accurate application of TDM. For this reason, the research activity has also been focused on first and second-generation antidepressant agents, like triciclic antidepressants, trazodone and m-chlorophenylpiperazine (m-cpp), paroxetine and its three main metabolites, venlafaxine and its active metabolite, and the most recent antidepressant introduced into the market, duloxetine. Among anxiolytics-hypnotics, benzodiazepines are very often involved in the pharmacotherapy of depression for the relief of anxious components; for this reason, it is useful to monitor these drugs, especially in cases of polypharmacy. The results obtained during these three years of Ph.D. program are reliable and the developed HPLC methods are suitable for the qualitative and quantitative determination of CNS drugs in biological fluids for TDM purposes.

Multi-level models and infrastructures for simulating biological system development

The hierarchical organisation of biological systems plays a crucial role in the pattern formation of gene expression resulting from the morphogenetic processes, where autonomous internal dynamics of cells, as well as cell-to-cell interactions through membranes, are responsible for the emergent peculiar structures of the individual phenotype. Being able to reproduce the systems dynamics at different levels of such a hierarchy might be very useful for studying such a complex phenomenon of self-organisation. The idea is to model the phenomenon in terms of a large and dynamic network of compartments, where the interplay between inter-compartment and intra-compartment events determines the emergent behaviour resulting in the formation of spatial patterns. According to these premises the thesis proposes a review of the different approaches already developed in modelling developmental biology problems, as well as the main models and infrastructures available in literature for modelling biological systems, analysing their capabilities in tackling multi-compartment / multi-level models. The thesis then introduces a practical framework, MS-BioNET, for modelling and simulating these scenarios exploiting the potential of multi-level dynamics. This is based on (i) a computational model featuring networks of compartments and an enhanced model of chemical reaction addressing molecule transfer, (ii) a logic-oriented language to flexibly specify complex simulation scenarios, and (iii) a simulation engine based on the many-species/many-channels optimised version of Gillespie’s direct method. The thesis finally proposes the adoption of the agent-based model as an approach capable of capture multi-level dynamics. To overcome the problem of parameter tuning in the model, the simulators are supplied with a module for parameter optimisation. The task is defined as an optimisation problem over the parameter space in which the objective function to be minimised is the distance between the output of the simulator and a target one. The problem is tackled with a metaheuristic algorithm. As an example of application of the MS-BioNET framework and of the agent-based model, a model of the first stages of Drosophila Melanogaster development is realised. The model goal is to generate the early spatial pattern of gap gene expression. The correctness of the models is shown comparing the simulation results with real data of gene expression with spatial and temporal resolution, acquired in free on-line sources.

A SMO inhibitor drug reduces the progenitor cell's maintenance in the Drosophila hematopoietic organ: a novel model to study Chronic Myelogenous Leukemia relapse

The chronic myeloid leukemia complexity and the difficulties of disease eradication have recently led to the development of drugs which, together with the inhibitors of TK, could eliminate leukemia stem cells preventing the occurrence of relapses in patients undergoing transplantation. The Hedgehog (Hh) signaling pathway positively regulates the self-renewal and the maintenance of leukemic stem cells and not, and this function is evolutionarily conserved. Using Drosophila as a model, we studied the efficacy of the SMO inhibitor drug that inhibit the human protein Smoothened (SMO). SMO is a crucial component in the signal transduction of Hh and its blockade in mammals leads to a reduction in the disease induction. Here we show that administration of the SMO inhibitor to animals has a specific effect directed against the Drosophila ortholog protein, causing loss of quiescence and hematopoietic precursors mobilization. The SMO inhibitor induces in L3 larvae the appearance of melanotic nodules generated as response by Drosophila immune system to the increase of its hemocytes. The same phenotype is induced even by the dsRNA:SMO specific expression in hematopoietic precursors of the lymph gland. The drug action is also confirmed at cellular level. The study of molecular markers has allowed us to demonstrate that SMO inhibitor leads to a reduction of the quiescent precursors and to an increase of the differentiated cells. Moreover administering the inhibitor to heterozygous for a null allele of Smo, we observe a significant increase in the phenotype penetrance compared to administration to wild type animals. This helps to confirm the specific effect of the drug itself. These data taken together indicate that the study of inhibitors of Smo in Drosophila can represent a useful way to dissect their action mechanism at the molecular-genetic level in order to collect information applicable to the studies of the disease in humans.

Decomposition Methods and Network Design Problems

Decomposition based approaches are recalled from primal and dual point of view. The possibility of building partially disaggregated reduced master problems is investigated. This extends the idea of aggregated-versus-disaggregated formulation to a gradual choice of alternative level of aggregation. Partial aggregation is applied to the linear multicommodity minimum cost flow problem. The possibility of having only partially aggregated bundles opens a wide range of alternatives with different trade-offs between the number of iterations and the required computation for solving it. This trade-off is explored for several sets of instances and the results are compared with the ones obtained by directly solving the natural node-arc formulation. An iterative solution process to the route assignment problem is proposed, based on the well-known Frank Wolfe algorithm. In order to provide a first feasible solution to the Frank Wolfe algorithm, a linear multicommodity min-cost flow problem is solved to optimality by using the decomposition techniques mentioned above. Solutions of this problem are useful for network orientation and design, especially in relation with public transportation systems as the Personal Rapid Transit. A single-commodity robust network design problem is addressed. In this, an undirected graph with edge costs is given together with a discrete set of balance matrices, representing different supply/demand scenarios. The goal is to determine the minimum cost installation of capacities on the edges such that the flow exchange is feasible for every scenario. A set of new instances that are computationally hard for the natural flow formulation are solved by means of a new heuristic algorithm. Finally, an efficient decomposition-based heuristic approach for a large scale stochastic unit commitment problem is presented. The addressed real-world stochastic problem employs at its core a deterministic unit commitment planning model developed by the California Independent System Operator (ISO).

Deep learning and embeddings for problems of computational biology

The development of Next Generation Sequencing promotes Biology in the Big Data era. The ever-increasing gap between proteins with known sequences and those with a complete functional annotation requires computational methods for automatic structure and functional annotation. My research has been focusing on proteins and led so far to the development of three novel tools, DeepREx, E-SNPs&GO and ISPRED-SEQ, based on Machine and Deep Learning approaches. DeepREx computes the solvent exposure of residues in a protein chain. This problem is relevant for the definition of structural constraints regarding the possible folding of the protein. DeepREx exploits Long Short-Term Memory layers to capture residue-level interactions between positions distant in the sequence, achieving state-of-the-art performances. With DeepRex, I conducted a large-scale analysis investigating the relationship between solvent exposure of a residue and its probability to be pathogenic upon mutation. E-SNPs&GO predicts the pathogenicity of a Single Residue Variation. Variations occurring on a protein sequence can have different effects, possibly leading to the onset of diseases. E-SNPs&GO exploits protein embeddings generated by two novel Protein Language Models (PLMs), as well as a new way of representing functional information coming from the Gene Ontology. The method achieves state-of-the-art performances and is extremely time-efficient when compared to traditional approaches. ISPRED-SEQ predicts the presence of Protein-Protein Interaction sites in a protein sequence. Knowing how a protein interacts with other molecules is crucial for accurate functional characterization. ISPRED-SEQ exploits a convolutional layer to parse local context after embedding the protein sequence with two novel PLMs, greatly surpassing the current state-of-the-art. All methods are published in international journals and are available as user-friendly web servers. They have been developed keeping in mind standard guidelines for FAIRness (FAIR: Findable, Accessible, Interoperable, Reusable) and are integrated into the public collection of tools provided by ELIXIR, the European infrastructure for Bioinformatics.