Due to their intrinsic nature, biological entities are universally considered as complex systems. Over years, many different computational methods pertaining to the Systems Biology field, have been devised to unravel this complexity. However, when taken alone, most of times these methods are not able to provide a deep comprehension of structural, spatial and dynamical aspects of the systems under evaluation. For this reason, approaches exploiting different levels of analysis are today a hot research topic in different areas, such as the theoretical formalization of the method, and the development of computational tools for the integration of different modeling perspectives. In the present dissertation I developed a computational pipeline able to perform analyses exploiting, one after the other, the three main modeling frameworks for biological systems, gaining, from every level, a different type of information: i.e. identification of flux distributions and metabolic sub-phenotypes from the ensemble evolutionary FBA (a novel method inspired by Flux Balance Analysis); information on network structural properties and topological metrics from graph theory approaches; estimation of kinetic constants for mechanism-based modeling through the definition of an efficient version of the Particle Swarm Optimizer based on Fuzzy Logic. Moreover, I also redefined a network visualization strategy able to overlay flux values and topological metrics to network structure. In order to validate the proposed pipeline I also developed a “core model” of yeast metabolism from which I identified two ensembles of flux distributions (possible solutions) in agreement with the “Crabtree-positive” and “Crabtree-negative” metabolic phenotypes. Moreover, by means of a cluster analysis, devised methods were able to define groups inside each ensemble that I identified as putative “sub-phenotypes”. Lastly, I contributed to reconstruct four reduced metabolic “core models”, deriving from the Human Metabolic Atlas, and describing three tissue-specific cancer conditions and a reference state. From these models a relevant heterogeneity emerged between reference and cancer conditions in terms of metabolic flux values.
A causa della loro natura intrinseca, le entità biologiche sono universalmente considerate sistemi complessi. Per dirimere tale complessità, nel corso degli anni, sono stati sviluppati innumerevoli metodi computazionali appartenenti all’ambito della Systems Biology. Tuttavia, quando utilizzati da soli, questi metodi non sono solitamente sufficienti per fornire un’approfondita comprensione degli aspetti strutturali, spaziali e dinamici dei sistemi studiati. Per questa ragione, approcci che coinvolgono diversi livelli di analisi sono oggi un promettente ambito di ricerca sotto diversi aspetti quali la formalizzazione del metodo e lo sviluppo di strumenti computazionali per l’integrazione delle diverse prospettive di modellazione. Nel mio lavoro di tesi, ho sviluppato una pipeline computazionale in grado di svolgere analisi che sfruttano, l’uno dopo l’altro, i tre principali framework di modellazione dei sistemi biologici, ottenendo da ogni livello un diverso tipo di informazione. In particolare sono state identificate distribuzioni di flusso e sotto-fenotipi metabolici attraverso la “Ensemble Evolutionary FBA” (un metodo innovativo inspirato dalla Flux Balance Analysis); informazioni su metriche topologiche e proprietà strutturali della rete da approcci di graph theory; stima di costanti cinetiche per la modellazione meccanicistica attraverso una versione efficiente del Particle Swarm Optimizer basato sulla Logica Fuzzy. Inoltre, ho anche ridefinito una strategia di visualizzazione della rete metabolica in grado di sovraimporre il valore dei flussi e le metriche topologiche alla struttura del network. Al fine di validare la pipeline ho sviluppato un “core model” del metabolismo di lievito, grazie al quale ho identificato due “ensemble” di distribuzioni di flusso (possibili soluzioni) in accordo con i fenotipi metabolici “Crabtree-positive" e “Crabtree-negative". Inoltre, attraverso i metodi sviluppati, è stato possibile isolare dei sottogruppi all’interno di ogni ensemble. Tali sottogruppi sono identificabili come sotto-fenotipi. Infine, ho contribuito alla ricostruzione di quattro “core models” metabolici ottenuti dalla riduzione di modelli “genome wide” (Human Metabolic Atlas) che illustrano tre condizioni tumorali tessuto specifiche e uno stato di riferimento. Da questi modelli è emersa una rilevante eterogeneità, in termini del valore dei flussi metabolici, tra la condizione di riferimento e i modelli tumorali.
(2015). A Computational Approach for Multi-Level Biological Complex Systems Analysis. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2015).
A Computational Approach for Multi-Level Biological Complex Systems Analysis
COLOMBO, RICCARDO
2015
Abstract
Due to their intrinsic nature, biological entities are universally considered as complex systems. Over years, many different computational methods pertaining to the Systems Biology field, have been devised to unravel this complexity. However, when taken alone, most of times these methods are not able to provide a deep comprehension of structural, spatial and dynamical aspects of the systems under evaluation. For this reason, approaches exploiting different levels of analysis are today a hot research topic in different areas, such as the theoretical formalization of the method, and the development of computational tools for the integration of different modeling perspectives. In the present dissertation I developed a computational pipeline able to perform analyses exploiting, one after the other, the three main modeling frameworks for biological systems, gaining, from every level, a different type of information: i.e. identification of flux distributions and metabolic sub-phenotypes from the ensemble evolutionary FBA (a novel method inspired by Flux Balance Analysis); information on network structural properties and topological metrics from graph theory approaches; estimation of kinetic constants for mechanism-based modeling through the definition of an efficient version of the Particle Swarm Optimizer based on Fuzzy Logic. Moreover, I also redefined a network visualization strategy able to overlay flux values and topological metrics to network structure. In order to validate the proposed pipeline I also developed a “core model” of yeast metabolism from which I identified two ensembles of flux distributions (possible solutions) in agreement with the “Crabtree-positive” and “Crabtree-negative” metabolic phenotypes. Moreover, by means of a cluster analysis, devised methods were able to define groups inside each ensemble that I identified as putative “sub-phenotypes”. Lastly, I contributed to reconstruct four reduced metabolic “core models”, deriving from the Human Metabolic Atlas, and describing three tissue-specific cancer conditions and a reference state. From these models a relevant heterogeneity emerged between reference and cancer conditions in terms of metabolic flux values.File | Dimensione | Formato | |
---|---|---|---|
PhD_unimib_760860.pdf
accesso aperto
Descrizione: Tesi dottorato
Tipologia di allegato:
Doctoral thesis
Dimensione
9.1 MB
Formato
Adobe PDF
|
9.1 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.