An inferential framework for biological network hypothesis tests

摘要: Networks are ubiquitous in modern cell biology and physiology. A large literature exists for inferring/proposing biological pathways/networks using statistical or machine learning algorithms. Despite these advances a formal testing procedure analyzing network-level observations is need of further development. Comparing the behaviour pharmacologically altered pathway to its canonical form an example salient one-sample comparison. Locating which pathways differentiate disease from no-disease phenotype may be recast as two-sample network inference problem. We outline inferential method performing one- hypothesis tests where sampling unit hypotheses stated via model(s). propose dissimilarity measure that incorporates nearby neighbour information contrast one more networks test. demonstrate explore utility our approach with both simulated microarray data; random graphs weighted (partial) correlation used models. Using well-known diabetes dataset ovarian cancer dataset, methods outlined here could better elucidate co-regulation changes between two clinically relevant phenotypes. Formal gene- protein-based logical progression existing gene-based gene-set differential expression. Commensurate growing appreciation development systems biology, dissimilarity-based presented allow us improve understanding other complex regulatory systems. The benefit was illustrated under select scenarios.