Computational inference of effective drugs for cancer cachexia treatment

Abstract

Drug repositioning is the application of known drugs and compound to new indications. Through the discovery of drug repositioning, there are many benefits, for example, reduced time, cost and risk compared with traditional drug development. Network-based approaches for drug repositioning have the advantage of discovering the relationships between drugs and disease-related genes, in order to get newly appeared attributes at a network status. Analysis of the path in the network is core idea in network-based approaches. Some measure is suggested to examine paths in the network. The closest measure has the good estimate in AUROC of drug-disease closeness. However, problems exist in the closest measure. It judges only the closest path among many of the shortest paths. In this research, new measure is suggested, which are named as the multiple closest measure, considering the number of the same closest path length from drug targets to disease-related genes. As a case study, cancer cachexia is selected. Cancer cachexia is characterized by systemic inflammation, loss of muscle, white adipose tissue wasting, anorexia and unbalance of energy expenditure. There is no computational inference of effective drugs for cancer cachexia treatment yet. I compare the performance between the closest measure and the multiple closest. The multiple closest measure has better performance than the closest measure. Through the multiple closest measure, I discover candidate drugs for cancer cachexia treatment. The multiple closest measure is advanced approach to detect repositioning drug in the network-based approach.