Iterative signature algorithm for the analysis of large-scale gene expression data.

摘要: We present an approach for the analysis of genome-wide expression data. Our method is designed to overcome limitations traditional techniques, when applied large-scale Rather than alloting each gene a single cluster, we assign both genes and conditions context-dependent potentially overlapping transcription modules. provide rigorous definition module as object be retrieved from An efficient algorithm, which searches modules encoded in data by iteratively refining sets until they match this definition, established. Each iteration involves linear map, induced normalized matrix, followed application threshold function. argue that our fact generalization singular value decomposition, corresponds special case where no applied. show analytically noisy leads better classification due implementation threshold. This result confirmed numerical analyses based on silico discuss briefly results obtained applying algorithm yeast Saccharomyces cerevisiae.