Search

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 condu

Different kinds of lesions can occur to DNA, and among them, one of the most dangerous is the double strand breaks (DSBs). Actually, DSBs can result in mutations, chromosome translocation or deletion. For this kind of lesions, depending on cell cycle phase as well as DNA-end resection, cells have developed specific repair pathways. Among these the error-free homologous recombination (HR) plays a crucial role. HR takes place during S/G2 phases, since the sister chromatids can be used as homologous templates. In this process, hRAD51 and BRCA2 are key players. hRAD51 is a recombinase of 339 amino-acids highly conserved through evolution which displays an intrinsic tendency to form oligomeric structures. BRCA2 is a very large protein of 3418 amino-acids, essential for the recruitment and accumulation of hRAD51 in the nucleus repairing-foci. BRCA2 interacts with hRAD51 through eight, so-called, BRC repeats, composed of 35-40 amino-acids. Mutations within this region have been linked to an increased risk of ovarian cancer development. In particular, several reports highlighted that missense mutations within one BRC repeat can hamper BRCA2 activity. Considering the close homology between the BRC repeats, it is striking how these mutations cannot be counterbalanced by the other non-mutated repeats preserving the function and the interactions of BRCA2 with hRAD51. To date the only interaction that has been structurally elucidated, is the one taking place amid the fourth BRC repeat and hRAD51. Only very little biophysical information is available on the interaction of the other BRC repeats with hRAD51. This thesis aims at elucidating the mechanism of hRAD51-BRCA2 interaction, by means of biophysical and structural approaches., Il nostro DNA è continuamente danneggiato da fattori endogeni ed esogeni. Ciò può determinare differenti tipologie di danno, tra le quali le rotture a doppio filamento sono considerabili come le più pericolose. Fortunatamente, le nostre cellule hanno sviluppato differenti strategie per riparare questi danni a seconda della fase del ciclo cellulare in cui si trovano. Tra questi meccanismi di riparazione del DNA un ruolo fondamentale è certamente giocato dalla cosiddetta ricombinazione omologa. Tale processo avviene nelle fasi S e G2 del ciclo cellulare e consente una riparazione del DNA molto accurata. In questo meccanismo di riparazione hRAD51 e BRCA2 sono due proteine fondamentali. hRAD51 è una recombinasi di 339 aminoacidi altamente conservata in diversi organismi, che possiede una tendenza intrinseca a formare strutture oligomeriche. BRCA2 è invece una proteina di notevoli dimensioni, 3418 aminoacidi, che è essenziale per il reclutamento e l’accumulo di hRAD51 nel nucleo. In particolare, si è dimostrato che BRCA2 interagisca con hRAD51 attraverso otto sequenze ripetute composte da 35-40 aminoacidi denominate “BRC-repeats”. Studi epidemiologici hanno evidenziato come le mutazioni all’interno di questa regione siano correlate ad un incremento del rischio di sviluppare cancro ovarico. Inoltre, è stato dimostrato come mutazioni che avvengono all’interno di una sola ripetizione BRC possano essere sufficienti a bloccare il reclutamento di hRAD51 e conseguentemente la ricombinazione omologa. Considerata l’alta omologia tra queste sequenze è sorprendente come tali mutazioni all’interno di una sola ripetizione BRC non possano essere controbilanciate dalle altre sette. Ad oggi, l’unica interazione che è stata caratterizzata dal punto di vista strutturale è quella che sussiste tra la quarta ripetizione BRC e hRAD51. Lo scopo di questa tesi è quello di studiare il meccanismo di interazione tra i “BRC repeats” e hRAD51 attraverso metodiche biofisiche e di biologia strutturale

Infection with severe acute respiratory syndrome coronavirus 2 associates with diverse symptoms, which can persist for months. While antiviral antibodies are protective, those targeting interferons and other immune factors are associated with adverse coronavirus disease 2019 (COVID-19) outcomes. Here we discovered that antibodies against specific chemokines were omnipresent post-COVID-19, were associated with favorable disease outcome and negatively correlated with the development of long COVID at 1 yr post-infection. Chemokine antibodies were also present in HIV-1 infection and autoimmune disorders, but they targeted different chemokines compared with COVID-19. Monoclonal antibodies derived from COVID-19 convalescents that bound to the chemokine N-loop impaired cell migration. Given the role of chemokines in orchestrating immune cell trafficking, naturally arising chemokine antibodies may modulate the inflammatory response and thus bear therapeutic potential.

Clinical and omics data are a promising field of application for machine learning techniques even though these methods are not yet systematically adopted in healthcare institutions. Despite artificial intelligence has proved successful in terms of prediction of pathologies or identification of their causes, the systematic adoption of these techniques still presents challenging issues due to the peculiarities of the analysed data. The aim of this thesis is to apply machine learning algorithms to both clinical and omics data sets in order to predict a patient's state of health and get better insights on the possible causes of the analysed diseases. In doing so, many of the arising issues when working with medical data will be discussed while possible solutions will be proposed to make machine learning provide feasible results and possibly become an effective and reliable support tool for healthcare systems.

In Cystic Fibrosis (CF) the deletion of phenylalanine 508 (F508del) in the CFTR anion channel is associated to misfolding and defective gating of the mutant protein. Among the known proteins involved in CFTR processing, one of the most promising drug target is the ubiquitin ligase RNF5, which normally promotes F508del-CFTR degradation. In this context, a small molecule RNF5 inhibitor is expected to chemically mimic a condition of RNF5 silencing, thus preventing mutant CFTR degradation and causing its stabilization and plasma membrane trafficking. Hence, by exploiting a virtual screening (VS) campaign, the hit compound inh-2 was discovered as the first-in-class inhibitor of RNF5. Evaluation of inh-2 efficacy on CFTR rescue showed that it efficiently decreases ubiquitination of mutant CFTR and increases chloride current in human primary bronchial epithelia. Based on the promising biological results obtained with inh-2, this thesis reports the structure-based design of potential RNF5 inhibitors having improved potency and efficacy. The optimization of general synthetic strategies gave access to a library of analogues of the 1,2,4-thiadiazol-5-ylidene inh-2 for SAR investigation. The new analogues were tested for their corrector activity in CFBE41o- cells by using the microfluorimetric HS-YFP assay as a primary screen. Then, the effect of putative RNF5 inhibitors on proliferation, apoptosis and the formation of autophagic vacuoles was evaluated. Some of the new analogs significantly increased the basal level of autophagy, reproducing RNF5 silencing effect in cell. Among them, one compound also displayed a greater rescue of the F508del-CFTR trafficking defect than inh-2. Our preliminary results suggest that the 1,2,4-thiadiazolylidene could be a suitable scaffold for the discovery of potential RNF5 inhibitors able to rescue mutant CFTRs. Biological tests are still ongoing to acquire in-depth knowledge about the mechanism of action and therapeutic relevance of this unprec

The ribosome stalling mechanism is a crucial biological process, yet its atomistic underpinning is still elusive. In this framework, the human XBP1u translational arrest peptide (AP) plays a central role in regulating the unfolded protein response (UPR) in eukaryotic cells. Here, we report multimicrosecond all-atom molecular dynamics simulations designed to probe the interactions between the XBP1u AP and the mammalian ribosome exit tunnel, both for the wild type AP and for four mutant variants of different arrest potencies. Enhanced sampling simulations allow investigating the AP release process of the different variants, shedding light on this complex mechanism. The present outcomes are in qualitative/quantitative agreement with available experimental data. In conclusion, we provide an unprecedented atomistic picture of this biological process and clear-cut insights into the key AP-ribosome interactions.

Infection by SARS-CoV-2 leads to diverse symptoms, which can persist for months. While antiviral antibodies are protective, those targeting interferons and other immune factors are associated with adverse COVID-19 outcomes. Instead, we discovered that antibodies against specific chemokines are omnipresent after COVID-19, associated with favorable disease, and predictive of lack of long COVID symptoms at one year post infection. Anti-chemokine antibodies are present also in HIV-1 and autoimmune disorders, but they target different chemokines than those in COVID-19. Finally, monoclonal antibodies derived from COVID- 19 convalescents that bind to the chemokine N-loop impair cell migration. Given the role of chemokines in orchestrating immune cell trafficking, naturally arising anti-chemokine antibodies associated with favorable COVID-19 may be beneficial by modulating the inflammatory response and thus bear therapeutic potential. One-sentence summary: Naturally arising anti-chemokine antibodies associate with favorable COVID-19 and are predictive of lack of long COVID.