Generate networks with power-law and exponential-law distributed degrees: with applications in link prediction of tumor pathways

摘要

In present study I proposed a method for generating biological networks based on power-law (p(x)=x-?) andexponential-law (p(x)=e-?x) distribution functions. Given the parameter of power-law or exponential-lawdistribution function, ?, the algorithm generates an expected frequency distribution according to the givenparameter, thereafter creates an adjacency matrix in which (practical) frequency distribution of node degreesmatches the expected frequency distribution. The results showed that power-law distribution function performsmuch better than exponential-law distribution function in generating networks. Using the revised algorithm,tumor related networks (pathways) are simulated and predicted. The results prove that the algorithm is overalleffective in predicting network links (14.6%?21.2% of correctly predicted links against 0.1%?3.4% of that forrandom assignments). Matlab codes of the algorithms are given also.