Dear Bioconductor users, Since version 11, the MBNI (Molecular & Behavioral Neuroscience Institute, University of Michigan) Custom CDF repository is added into Bioconductor. Therefore, Bioconductor user can use biocLite() to install those R packages, while the traditional Download/R CMD INSTALL {} still works. The custom CDF is designed to make the GeneChip probe set definitions consistent with the latest version of the genome and transcriptome databases. Several systematic analyses show that the updated probe set definition can lead to 30%-50% differences in the list of differentially expressed genes and more significant changes in gene set-based analysis approaches. We currently support most Affymetrix GeneChips and generate probe set definitions based on several major gene and transcript definitions for each species. The following is a list of commonly used custom CDFs and their pros and cons: 1. Entrez Gene based CDF: most widely used, excellent for gene-based target definitions. One probe set for one unique gene in the corresponding database. 2. Refseq-based CDF: most stable. good for transcript-based analysis. The shortcoming is probe sets representing different transcripts from the same gene may be identical or highly similar due to the lack of transcript-specific probes on GeneChip. 3. UniGene-based CDF: used to be the preferred choice if the goal is to represent as many genes as possible. We recommend Entrez Gene for species that have similar gene-based probe set count, which include human and mouse now. 4. ENSEMBL/VEGA Gene/Transcript/Exon: for researchers prefer the ENSEMBL/VEGA system. VEGA is supposed to be expert curated thus its gene/transcript/exon definitions are conceivably more accurate. The exon-based probe sets can be used to detect some alternative splicing events even in GeneChips not designed for splicing analysis. We want to thank James MacDonald, Marc Carlson and Patrick Aboyoun for helping us to setup custom CDFs on the bioconductor system. We also want to thank many users for their suggestions, which are essential for the continuous improvement of custom CDFs. Our work is supported by the Pritzker Neuropsychiatric Disorders Research Consortium and the National Center for Integrated Biomedical Informatics. Best, Manhong Dai Molecular & Behavioral Neuroscience Institute University of Michigan