With the advent of high-throughput wet lab technologies the amount of protein interaction data available publicly has increased substantially, in turn spurring a plethora of computational methods for in silico knowledge discovery from this data. In this paper, we focus on parameterized methods for modeling and solving complex computational problems encountered in such knowledge discovery from protein data. Specifically, we concentrate on three relevant problems today in proteomics, namely detection of lethal proteins, functional modules and alignments from protein interaction networks. We propose novel graph theoretic models for these problems and devise practical parameterized algorithms. At a broader level, we demonstrate how these methods can be viable alternatives for the several heurestic, randomized, approximation and sub-optimal methods by arriving at parameterized yet optimal solutions for these problems. We substantiate these theoretical results by experimenting on real protein interaction data of S. cerevisiae (budding yeast) and verifying the results using gene ontology.
تحميل البحث