Systems Biology encompasses different research areas, sharing graph theory as a common conceptual framework. Its main focus is the modelling and investigation of molecular interactions as complex networks. Notably, although experimental datasets allow the construction of context-specific molecular networks, the effect of quantitative variations of molecular states, i.e. the biochemical status, is not in- corporated into the current network topologies. This fact poses great limitations in terms of predictive power. To overcome these limitations we have developed a novel methodology that allows incorporating experimental quantitative data into the graph topology, thus leading to a potentiated network representation. It is now possible to model, at graph level, the outcome of a specific experimental analysis. The mathematical approach, based on a demonstrated theorem, was validated in four different pathological contexts, including B-Cell Lymphocytic Leukaemia, Amyloidosis, Pancreatic Endocrine Tumours and Myocardial Infarc- tion. Reconstructing disease-specific, potentiated networks coupled to topolog- ical analysis and machine learning techniques allowed the automatic discrimina- tion of healthy versus unhealthy subjects in every context. Our methodology takes advantage of the topological information extracted from protein-protein in- teractions networks integrating experimental data into their topology. Incorpo- rating quantitative data of molecular state into graphs permits to obtain enriched representations that are tailored to a specific experimental condition, or to a sub- ject, leading to an effective personalised approach. Moreover, in order to validate the biological results, we have developed an app, for the Cytoscape platform, that allows the creation of randomised networks and the randomisation of exist- ing, real networks. Since there is a lack of tools for generating and randomising networks, our app helps researchers to exploit different, well known random net- work models that could be used as a benchmark for validating the outcomes from real datasets. We also proposed three possibile approaches for creating randomly weighted networks starting from the experimental, quantitative data. Finally, some of the functionalities of our app, plus some other functions, were devel- oped, in R, to allow exploiting the potential of this language and to perform network analysis using our multiplication model. In summary, we developed a workflow that starts from the creation of a set of personalised networks that are able to integrate numerical information. We gave some directions that guide the researchers in performing the network analysis. Finally, we developed a Java App and some R functions that permit to validate all the findings using a random network based approach.
Biological network analysis: from topological indexes to biological applications towards personalised medicine.
Tosadori, Gabriele
2017-01-01
Abstract
Systems Biology encompasses different research areas, sharing graph theory as a common conceptual framework. Its main focus is the modelling and investigation of molecular interactions as complex networks. Notably, although experimental datasets allow the construction of context-specific molecular networks, the effect of quantitative variations of molecular states, i.e. the biochemical status, is not in- corporated into the current network topologies. This fact poses great limitations in terms of predictive power. To overcome these limitations we have developed a novel methodology that allows incorporating experimental quantitative data into the graph topology, thus leading to a potentiated network representation. It is now possible to model, at graph level, the outcome of a specific experimental analysis. The mathematical approach, based on a demonstrated theorem, was validated in four different pathological contexts, including B-Cell Lymphocytic Leukaemia, Amyloidosis, Pancreatic Endocrine Tumours and Myocardial Infarc- tion. Reconstructing disease-specific, potentiated networks coupled to topolog- ical analysis and machine learning techniques allowed the automatic discrimina- tion of healthy versus unhealthy subjects in every context. Our methodology takes advantage of the topological information extracted from protein-protein in- teractions networks integrating experimental data into their topology. Incorpo- rating quantitative data of molecular state into graphs permits to obtain enriched representations that are tailored to a specific experimental condition, or to a sub- ject, leading to an effective personalised approach. Moreover, in order to validate the biological results, we have developed an app, for the Cytoscape platform, that allows the creation of randomised networks and the randomisation of exist- ing, real networks. Since there is a lack of tools for generating and randomising networks, our app helps researchers to exploit different, well known random net- work models that could be used as a benchmark for validating the outcomes from real datasets. We also proposed three possibile approaches for creating randomly weighted networks starting from the experimental, quantitative data. Finally, some of the functionalities of our app, plus some other functions, were devel- oped, in R, to allow exploiting the potential of this language and to perform network analysis using our multiplication model. In summary, we developed a workflow that starts from the creation of a set of personalised networks that are able to integrate numerical information. We gave some directions that guide the researchers in performing the network analysis. Finally, we developed a Java App and some R functions that permit to validate all the findings using a random network based approach.File | Dimensione | Formato | |
---|---|---|---|
Gabriele_Tosadori_Doctoral_Thesis.pdf
accesso aperto
Licenza:
Creative commons
Dimensione
2.67 MB
Formato
Adobe PDF
|
2.67 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.