Pre-Processing Optimization of RNA Immunoprecipitation Microarray Data

Abstract

Pre-mRNA splicing is an essential step in the post-transcriptional gene expression control involving protein-splicing factors like U2AF, which is exported to the cytoplasm and implicated in additional cellular functions. Identification of U2AF-associated mRNAs under native conditions was performed by immunoprecipitation and hybridization to Affymetrix GeneChip. Normalization and gene selection methods were performed, but the results were not reliable as they were different for different procedures, mainly because more than 20% of the mRNAs detected are differently enriched and the common normalization methods are based on small differences between them. We implemented a background correction method inspired in a non-specific hybridization method used for pre-processing data from ChIP-Chip technology. In this work, linear regression models are used to model in each array the non-specific hybridization, accounting for interactions between each three consecutive nucleotides into the probe sequence. Every probe intensity on the array was standardized using its predicted intensity and the probes' variance for similar predicted intensities. The standardized probe intensity values showed no need for further normalization and could be directly compared. We propose a probe set score, and a probe set enrichment value (ENRval) and its respective p-value for gene enrichment selection.