On parallelizing graph theoretical approaches for identifying causal genes and pathways from very large biological networks

The time taken to compute meaningful information from biological networks is very high for most of the applications and therefore these networks are very hard to process. This work focuses on improving the speed of processing biological networks, in particular, faster traversal of genomes which have been mapped into a network for the detection of causal genes and associated pathways. Inference of disease causing genes and their pathways has achieved a crucial role in computational biology because of its practicality in understanding the major causal genes and their interactions that lead to a disease state, and suggesting new drug targets. In this work, Hadoop's distributed storage system has been used to store the molecular interaction network. Graph parallel processing techniques of Hadoop MapReduce, in conjunction with graph theoretical approaches have been utilized to improve the accuracy of results and execution time on benchmark data.