UC San Diego

Community Detection is an interesting computational technique for the analysis of networks. This technique can yield useful insights into the structural organization of a network, and can serve as a basis for understanding the correspondence between structure and function (specific to the domain of the network). In this dissertation, I have sought to leverage this technique for the study of biological networks of practical relevance and significance.

The study begins with an exploration of existing techniques for Community Detection, following which an optimization is proposed for one of the widely used graph-theoretic approaches. As the next step, an investigation is performed on the suitability of a machine-learning based algorithm for Community Detection in the context of biological networks. Subsequently, the use of Community Detection for understanding pathology with a specific focus on Duchenne Muscular Dystrophy (DMD), is explored. This illustrated key distinguishing features in the structural and functional organization of the constituent biological pathways as it relates to DMD. Finally, a novel algorithm for Community Detection is proposed, which is motivated by a physical systems analogy. An analysis of the algorithm's properties, together with its applications to biological networks, is also presented.

I believe that the techniques and algorithms developed as part of this dissertation in the context of biological networks, have the potential to open up new vistas for therapeutic applications such as targeted drug development.