Using Smoothing Splines to Select Significant Genes in Microarrays

Abstract

<p>DNA microarray technology has been widely used in many applications such as gene discovery, disease research and drug investigation. This thesis is based on a project studying the genetics of Type 1 Diabetes.</p> <p>In this thesis we introduce a method to use smoothing splines to select significant genes in microarrays. This method is based on significance analysis of microarrays (SAM). We choose upper and lower significance cut-offs based on when the numerical derivative of the spline exceeds a threshold. We declare that any genes whose observed statistics are less than the lower cut-off or greater than the upper cut-off to be significant. We also explain how to use this method to calculate the number of significant genes and estimate false discovery rates.</p> <p>We use both Affymetrix and Illumina real data sets in our analysis and the results are satisfactory. We try to use the simulation study to test our method but we have a problem that we can not generate simulated data which is similar to the real microarray data.</p>