Correlation Vectors in Gene Expression Profiling Modern methods of microarray data analysis are biased towards selecting those genes that display the most pronounced differential expression. Our preliminary studies strongly suggest that the correlation structure of microarray data can serve as an important indicator of changes in transcription profiles between phenotypes. The objective of the proposed project is to develop a new multivariate method for selecting candidate genes that change their relationships (correlations) with other genes across the conditions (phenotypes) under study. This will be accomplished by designing a pertinent multivariate statistical test and associated resampling algorithm. The proposed method represents a radical conceptual change in current approaches focused solely on differentially expressed genes. This new approach will enable the biologist to enrich currently practiced methods of microarray data analysis with quantitative inference on dependencies between gene expression intensities. An efficient multiple testing rule will be designed to combine the two approaches in a single procedure that provides strong control of false discoveries. The proposed general methodology will provide biological investigators with an additional source of information, allowing them to make more educated decisions when selecting and prioritizing candidate genes from microarray data analysis. This methodology has far-reaching implications for diverse research activities in modern genomics that involve uncovering molecular mechanisms of diseases and finding therapeutic targets. ? ? ?