Machine learning tools for protein annotation: the cases of transmembrane β-barrel and myristoylated proteins

Biology is now a “Big Data Science” thanks to technological advancements allowing the characterization of the whole macromolecular content of a cell or a collection of cells. This opens interesting perspectives, but only a small portion of this data may be experimentally characterized. From this derives the demand of accurate and efficient computational tools for automatic annotation of biological molecules. This is even more true when dealing with membrane proteins, on which my research project is focused leading to the development of two machine learning-based methods: BetAware-Deep and SVMyr. BetAware-Deep is a tool for the detection and topology prediction of transmembrane beta-barrel proteins found in Gram-negative bacteria. These proteins are involved in many biological processes and primary candidates as drug targets. BetAware-Deep exploits the combination of a deep learning framework (bidirectional long short-term memory) and a probabilistic graphical model (grammatical-restrained hidden conditional random field). Moreover, it introduced a modified formulation of the hydrophobic moment, designed to include the evolutionary information. BetAware-Deep outperformed all the available methods in topology prediction and reported high scores in the detection task. Glycine myristoylation in Eukaryotes is the binding of a myristic acid on an N-terminal glycine. SVMyr is a fast method based on support vector machines designed to predict this modification in dataset of proteomic scale. It uses as input octapeptides and exploits computational scores derived from experimental examples and mean physicochemical features. SVMyr outperformed all the available methods for co-translational myristoylation prediction. In addition, it allows (as a unique feature) the prediction of post-translational myristoylation. Both the tools here described are designed having in mind best practices for the development of machine learning-based tools outlined by the bioinformatics community. Moreover, they are made available via user-friendly web servers. All this make them valuable tools for filling the gap between sequential and annotated data.

Organizzazione strutturale della catena respiratoria mitocondriale

La catena respiratoria mitocondriale è principalmente costituita da proteine integrali della membrana interna, che hanno la capacità di accoppiare il flusso elettronico, dovuto alle reazioni redox che esse catalizzano, al trasporto di protoni dalla matrice del mitocondrio verso lo spazio intermembrana. Qui i protoni accumulati creano un gradiente elettrochimico utile per la sintesi di ATP ad opera dell’ATP sintasi. Nonostante i notevoli sviluppi della ricerca sulla struttura e sul meccanismo d’azione dei singoli enzimi della catena, la sua organizzazione sovramolecolare, e le implicazioni funzionali che ne derivano, rimangono ancora da chiarire in maniera completa. Da questa problematica trae scopo la presente tesi volta allo studio dell’organizzazione strutturale sovramolecolare della catena respiratoria mediante indagini sia cinetiche che strutturali. Il modello di catena respiratoria più accreditato fino a qualche anno fa si basava sulla teoria delle collisioni casuali (random collision model) che considera i complessi come unità disperse nel doppio strato lipidico, ma collegate funzionalmente tra loro da componenti a basso peso molecolare (Coenzima Q10 e citocromo c). Recenti studi favoriscono invece una organizzazione almeno in parte in stato solido, in cui gli enzimi respiratori si presentano sotto forma di supercomplessi (respirosoma) con indirizzamento diretto (channeling) degli elettroni tra tutti i costituenti, senza distinzione tra fissi e mobili. L’importanza della comprensione delle relazioni che si instaurano tra i complessi , deriva dal fatto che la catena respiratoria gioca un ruolo fondamentale nell’invecchiamento, e nello sviluppo di alcune malattie cronico degenerative attraverso la genesi di specie reattive dell’ossigeno (ROS). E’ noto, infatti, che i ROS aggrediscono, anche i complessi respiratori e che questi, danneggiati, producono più ROS per cui si instaura un circolo vizioso difficile da interrompere. La nostra ipotesi è che, oltre al danno a carico dei singoli complessi, esista una correlazione tra le modificazioni della struttura del supercomplesso, stress ossidativo e deficit energetico. Infatti, la dissociazione del supercomplesso può influenzare la stabilità del Complesso I ed avere ripercussioni sul trasferimento elettronico e protonico; per cui non si può escludere che ciò porti ad un’ulteriore produzione di specie reattive dell’ossigeno. I dati sperimentali prodotti a sostegno del modello del respirosoma si riferiscono principalmente a studi strutturali di elettroforesi su gel di poliacrilammide in condizioni non denaturanti (BN-PAGE) che, però, non danno alcuna informazione sulla funzionalità dei supercomplessi. Pertanto nel nostro laboratorio, abbiamo sviluppato una indagine di tipo cinetico, basata sull’analisi del controllo di flusso metabolico,in grado di distinguere, funzionalmente, tra supercomplessi e complessi respiratori separati. Ciò è possibile in quanto, secondo la teoria del controllo di flusso, in un percorso metabolico lineare composto da una serie di enzimi distinti e connessi da intermedi mobili, ciascun enzima esercita un controllo (percentuale) differente sull’intero flusso metabolico; tale controllo è definito dal coefficiente di controllo di flusso, e la somma di tutti i coefficienti è uguale a 1. In un supercomplesso, invece, gli enzimi sono organizzati come subunità di una entità singola. In questo modo, ognuno di essi controlla in maniera esclusiva l’intero flusso metabolico e mostra un coefficiente di controllo di flusso pari a 1 per cui la somma dei coefficienti di tutti gli elementi del supercomplesso sarà maggiore di 1. In questa tesi sono riportati i risultati dell’analisi cinetica condotta su mitocondri di fegato di ratto (RLM) sia disaccoppiati, che accoppiati in condizioni fosforilanti (stato 3) e non fosforilanti (stato 4). L’analisi ha evidenziato l’associazione preferenziale del Complesso I e Complesso III sia in mitocondri disaccoppiati che accoppiati in stato 3 di respirazione. Quest’ultimo risultato permette per la prima volta di affermare che il supercomplesso I+III è presente anche in mitocondri integri capaci della fosforilazione ossidativa e che il trasferimento elettronico tra i due complessi possa effettivamente realizzarsi anche in condizioni fisiologiche, attraverso un fenomeno di channeling del Coenzima Q10. Sugli stessi campioni è stata eseguita anche un analisi strutturale mediante gel-elettroforesi (2D BN/SDS-PAGE) ed immunoblotting che, oltre a supportare i dati cinetici sullo stato di aggregazione dei complessi respiratori, ci ha permesso di evidenziare il ruolo del citocromo c nel supercomplesso, in particolare per il Complesso IV e di avviare uno studio comparativo esteso ai mitocondri di cuore bovino (BHM), di tubero di patata (POM) e di S. cerevisiae.

Essays in Applied Insurance Economics

The papers included in this thesis deal with a few aspects of insurance economics that have seldom been dealt with in the applied literature. In the first paper I apply for the first time the tools of the economics of crime to study the determinants of frauds, using data on Italian provinces. The contributions to the literature are manifold: -The price of insuring has a positive correlation with the propensity to defraud -Social norms constraint fraudulent behavior, but their strength is curtailed in economic downturns -I apply a simple extension of the Random Coefficient model, which allows for the presence of time invariant covariates and asymmetries in the impact of the regressors. The second paper assesses how the evolution of macro prudential regulation of insurance companies has been reflected in their equity price. I employ a standard event study methodology, deriving the definition of the “control” and “treatment” groups from what is implied by the regulatory framework. The main results are: -Markets care about the evolution of the legislation. Their perception has shifted from a first positive assessment of a possible implicit “too big to fail” subsidy to a more negative one related to its cost in terms of stricter capital requirement -The size of this phenomenon is positively related to leverage, size and on the geographical location of the insurance companies The third paper introduces a novel methodology to forecast non-life insurance premiums and profitability as function of macroeconomic variables, using the simultaneous equation framework traditionally employed macroeconometric models and a simple theoretical model of insurance pricing to derive a long term relationship between premiums, claims expenses and short term rates. The model is shown to provide a better forecast of premiums and profitability compared with the single equation specifications commonly used in applied analysis.

Frequency domain analysis of stationary time series

This thesis provides a necessary and sufficient condition for asymptotic efficiency of a nonparametric estimator of the generalised autocovariance function of a Gaussian stationary random process. The generalised autocovariance function is the inverse Fourier transform of a power transformation of the spectral density, and encompasses the traditional and inverse autocovariance functions. Its nonparametric estimator is based on the inverse discrete Fourier transform of the same power transformation of the pooled periodogram. The general result is then applied to the class of Gaussian stationary ARMA processes and its implications are discussed. We illustrate that for a class of contrast functionals and spectral densities, the minimum contrast estimator of the spectral density satisfies a Yule-Walker system of equations in the generalised autocovariance estimator. Selection of the pooling parameter, which characterizes the nonparametric estimator of the generalised autocovariance, controlling its resolution, is addressed by using a multiplicative periodogram bootstrap to estimate the finite-sample distribution of the estimator. A multivariate extension of recently introduced spectral models for univariate time series is considered, and an algorithm for the coefficients of a power transformation of matrix polynomials is derived, which allows to obtain the Wold coefficients from the matrix coefficients characterizing the generalised matrix cepstral models. This algorithm also allows the definition of the matrix variance profile, providing important quantities for vector time series analysis. A nonparametric estimator based on a transformation of the smoothed periodogram is proposed for estimation of the matrix variance profile.

Exploring the internal kinematics of Galactic Globular Cluster cores

Globular clusters (GCs) are traditionally described as simple quasi-relaxed non-rotating stellar systems, characterized by spherical symmetry and isotropy in velocity space. However, recent studies have shown deviations from isotropic velocity distributions and significant internal rotation in many GCs, suggesting that their internal structure and kinematics are more complex than previously thought. The aim of this thesis is to investigate the internal kinematics of Galactic Globular Clusters (GGCs) as part of the Multi-Instrument Kinematic Survey (MIKiS), which exploits the capabilities of different ESO-VLT spectrographs to obtain comprehensive velocity dispersion (VD) and rotation profiles of GGCs. Moreover, this thesis has the particular goal of unraveling the kinematics of GC cores, which are still largely unexplored, by taking advantage of the exceptional spatial resolution of the adaptive-optics assisted integral-field spectrograph MUSE/NFM. The thesis presents a thorough kinematic study of three GGCs NGC 1904, NGC 6440, and NGC 6569. By combining the data sets acquired with four different spectrographs, we obtained the radial velocity (RV) of more than 1000 individual stars in each cluster, sampling from the innermost to the outermost regions. This allowed us to obtain the entire VD profile of each cluster and exclude the presence of an intermediate-mass black hole in the core of NGC 1904, at odds with previous findings obtained from integrated-light spectra. The studies also revealed signatures of internal rotation in each of the GCs studied. These results, supported by those of N-body simulations, prove that GCs were born with a significant initial rotation that they gradually lost through internal two-body relaxation and angular momentum loss carried away by escaping stars. Furthermore, we derived the structural parameters of NGC 6440 and NGC 6569, obtaining a comprehensive overview of the internal kinematics and structure of these GCs, which is necessary to properly reconstruct the evolutionary history of these systems.

Analysis of earthquakes predictability

Earthquake prediction is a complex task for scientists due to the rare occurrence of high-intensity earthquakes and their inaccessible depths. Despite this challenge, it is a priority to protect infrastructure, and populations living in areas of high seismic risk. Reliable forecasting requires comprehensive knowledge of seismic phenomena. In this thesis, the development, application, and comparison of both deterministic and probabilistic forecasting methods is shown. Regarding the deterministic approach, the implementation of an alarm-based method using the occurrence of strong (fore)shocks, widely felt by the population, as a precursor signal is described. This model is then applied for retrospective prediction of Italian earthquakes of magnitude M≥5.0,5.5,6.0, occurred in Italy from 1960 to 2020. Retrospective performance testing is carried out using tests and statistics specific to deterministic alarm-based models. Regarding probabilistic models, this thesis focuses mainly on the EEPAS and ETAS models. Although the EEPAS model has been previously applied and tested in some regions of the world, it has never been used for forecasting Italian earthquakes. In the thesis, the EEPAS model is used to retrospectively forecast Italian shallow earthquakes with a magnitude of M≥5.0 using new MATLAB software. The forecasting performance of the probabilistic models was compared to other models using CSEP binary tests. The EEPAS and ETAS models showed different characteristics for forecasting Italian earthquakes, with EEPAS performing better in the long-term and ETAS performing better in the short-term. The FORE model based on strong precursor quakes is compared to EEPAS and ETAS using an alarm-based deterministic approach. All models perform better than a random forecasting model, with ETAS and FORE models showing better performance. However, to fully evaluate forecasting performance, prospective tests should be conducted. The lack of objective tests for evaluating deterministic models and comparing them with probabilistic ones was a challenge faced during the study.

Probabilistic approaches in civil engineering: generation of random fields and structural identification with genetic algorithms

The inherent stochastic character of most of the physical quantities involved in engineering models has led to an always increasing interest for probabilistic analysis. Many approaches to stochastic analysis have been proposed. However, it is widely acknowledged that the only universal method available to solve accurately any kind of stochastic mechanics problem is Monte Carlo Simulation. One of the key parts in the implementation of this technique is the accurate and efficient generation of samples of the random processes and fields involved in the problem at hand. In the present thesis an original method for the simulation of homogeneous, multi-dimensional, multi-variate, non-Gaussian random fields is proposed. The algorithm has proved to be very accurate in matching both the target spectrum and the marginal probability. The computational efficiency and robustness are very good too, even when dealing with strongly non-Gaussian distributions. What is more, the resulting samples posses all the relevant, welldefined and desired properties of “translation fields”, including crossing rates and distributions of extremes. The topic of the second part of the thesis lies in the field of non-destructive parametric structural identification. Its objective is to evaluate the mechanical characteristics of constituent bars in existing truss structures, using static loads and strain measurements. In the cases of missing data and of damages that interest only a small portion of the bar, Genetic Algorithm have proved to be an effective tool to solve the problem.