Discovery of overlapping 1-closed biclusters
Abhik Kumar Banerjee · OhioLink ETD Center (Ohio Library and Information Network) · 2012
The aim of this thesis is to find overlapping pairs of 1-closed biclusters from two different but related heterogeneous relations.Our algorithm generates 1-closed biclusters by expanding closed biclusters.In the process of generating 1-closed bicluster we allow to add limited "0"s to the closed bicluster.A bicluster with a limited number of such "0"s still captures the close link among the two sets and additionally, help predict the edges that may be missing in the original dataset.We have designed a novel algorithm, and tested it using a synthetic and two biomedical datasets from the field of genomics.We predict target disease -gene relationships which are relatively weakly linked as compared to a closed bicluster in the biomedical datasets.With the help of the synthetic dataset we also show that our algorithm is generic and can be applied to various fields including social networking and bioinformatics.iii