《基因芯片》.ppt

Key Features in the Design and Analysis of DNA Microarray Studies;/brb;Experimental Design;Class Prediction;Class Comparison;Outline of Presentation;Myth;Good microarray studies have clear objectives, but not generally gene specific mechanistic hypotheses Design and Analysis Methods Should Be Tailored to Study Objectives;Common Types of Objectives;Do Expression Profiles Differ for Defined Classes of Samples?;Myth;;Levels of Replication;;Allocation of Specimens toDual Label Arrays for Simple Class Comparison Problems ;Reference Design;;;Balanced Block Design;Loop Design;;Common Reference Designs;Loop Designs;Balanced Block Designs;Dye Bias;;cDNA experiment estimated sizes of the gene-specific dye bias for each of the 8,604 genes. An effect of size 1 corresponds to a 2-fold change in expression;Myth;Dye swap technical replicates of the same two rna samples are rarely necessary;Common Reference Design;Direct Comparison Designs;Sample Size Planning;Single Label ArraysComparing 2 equal size classes ;;Dual Label Arrays With Reference DesignComparing 2 equal size classes ;;?=0.001 ?=0.05 ?=1 ?2+2?2=0.25, ?2/?2=4human tumors;Controlling Expected False Discovery Rate;Can I reduce the number of arrays by pooling specimens?;Dual Label Arrays With Reference DesignPools of k Biological Samples ;Number of arrays and samples required for various pooling levels. An independent pool is constructed for each array, so that no sample is represented on more than one array. ?=0.001, ?=0.05, ?=1, ?2=0.384, ?2=0.054, m=1 ;Avoid Confounding;Components of Class Prediction;Feature Selection;Linear Classifiers for Two Classes;Linear Classifiers for Two Classes;Linear Classifiers for Two Classes;When pn;Myth;;Evaluating a Classifier;Split-Sample Evaluation;Leave-one-out Cross Validation;Leave-one-out Cross Validation;;Prediction on Simulated Null Data;;Permutation Distribution of Cross-validated Misclassification Rate of a Multivariate Classifier;Gene-Expression Profiles in Hereditary