Methods for comparing transcriptional patterns to determine potential clues to biological function

Abstract

Human mitosis is both activated and accompanied by genome-wide transcriptional changes, suggesting regulation of diverse biological pathways. Cancer-associated genetic alterations that disrupt the cell cycle often affect these broad transcriptional patterns, identifying discrete biological processes potentially involved in tumour-associated cellular phenotypes. To systematically identify biological processes that may be activated during the human cell cycle, we characterized more than 42,000 transcript levels during mitosis in fibroblasts. To study how tumorigenic conditions might affect a subset of these functions, we also performed this analysis on fibroblasts deficient in retinoblastoma gene product function. To analyse this data, we filtered the data based on the confidence level, clustered the remaining genes and have developed the first algorithms for statistically linking genome-wide transcriptional patterns and biological function. These methods clearly demonstrate association of function with specific transcriptional patterns, and provide means to automatically compare transcriptional patterns to determine potential clues to biological function.