Poster abstracts 78 Teichmann, Ulrike Development of a mouse-specific cDNA set for microarray gene expression analysis: differential gene expression in melanocyte lineage samples of the neural crest U. Teichmann 1 , G. Schuler 4 , T. M oore 5 , T. Otsuka 3 , G. M erlino 3 , J. Khan 2 , Y. Chen 2 , M . Bittner 2 , P. M eltzer 2 , J. Trent 2 & W. Pavan 1 1 Genetic Disease Research Branch, 2 Cancer Genetics Branch, National Human Genome Research Institute, 3 Laboratory of Molecular Biology, National Cancer Institute, 4 National Center for Biotechnology Information, National Institutes of Health, Bethesda, Maryland 20892, USA 5 Research Genetics, Huntsville, Alabama 35801, USA We are developing sets of mouse cDNAs with low redundancy for use in microarray analysis. The sets are based on sequence information from GenBank and the Washington University/Merck EST sequencing project. Related sequences from the non-redundant GenBank database were compared with mouse EST clones from the IMAGE consortium. A single EST for a given clus- ter was selected, colony purified and sequence validated. To evaluate this set of cDNAs for its use in differential gene expression studies, we generated cDNA microarrays. Those microarray slides were used in hybridization experiments using samples of the melanocyte-specific lineage of the neural crest as a model system. Gene expression was compared between RNA samples obtained from mouse melanoma tissue samples of a transgenic hepatocyte growth factor/scat- ter factor melanoma mouse model and a MEF-derived mouse fibroblast cell line (NIH3T3). Applying a confidence level of 99%, 8.5% of the arrayed genes were found to be differentially expressed. The melanocyte lineage marker genes Dct, Pmel17 and Sox10 are overexpressed in the melanoma samples with a ratio of 7.6, 6.9 and 7.8, respectively. A low expression ratio of 0.6 is seen for the fibroblast marker gene fibronectin. Genes encoding ubiquitin and the heat shock protein Hsc73 display no significant expression changes between the two dif- ferential samples. Teng, Chi-Hse Experimental designs using Affymetrix GeneChips Chi-Hse Teng 1 , Ann Nestorowicz 2 & Anne Reifel-M iller 2 1 Statistical and Mathematical Sciences, Eli Lilly and Company, Indianapolis, Indiana, USA 2 Endocrine Research, Eli Lilly and Company, Indianapolis, Indiana, USA Affymetrix GeneChip has been a popular and power tool for detecting mRNA transcripts to monitor genome-wide gene expression. Several sources of vari- ability could have influence on the results, for example, different genetic back- ground, RNA sample pooling, labeling/hybridisation processes, and different chips. Statistical concepts, randomisation, replication, blocking, and averaging, are introduced. Five experimental designs providing different information are demonstrated. Design 1 provides information on chip-to-chip variability. Design 2 provides information on treatment effects. Design 3 provides informa- tion on treatment effects and probe/chip-to-probe/chip variability. Design 4 pro- vides information on treatment effects and individual/probe/chip-to-individ- ual/probe/chip variability. Design 5 provides information on treatment effects, individual-to-individual variability, and probe/chip-to-probe/chip variability. The strength and weakness of each design will be discussed. These design examples are intended to demonstrate the concepts rather than real scenarios. Different prior knowledge and assumptions will have an effect on the design choice. Theilhaber, Joachim Bayesian estimation of fold-changes in gene expression: the PFOLD algorithm and its uses in the analysis of complex expression profiles Joachim Theilhaber, Steven Bushnell & Rainer Fuchs Hoechst-ARIAD Genomics Center, 26 Landsdowne Street Cambridge, Massachusetts 02139, USA The relatively low signal-to-noise ratio prevalent in expression technology, typi- cally in the range 2-5 when “signal” is defined as the median level of gene expres- sion, makes quality metrics such as confidence intervals and statistical signifi- cance essential in interpreting the data. We present models for the several sources of noise which affect expression measurements, including cross-hybridisation and chip-to-chip variation, and discuss how the model parameters may be obtained on-the-fly, either from single-scan data or from replicate experiments. These noise models have been integrated into the PFOLD algorithm: this is a method, ground- ed in a Bayesian framework, which not only estimates fold-changes, but also assigns a statistical significance to the change through a P-value. We have found that the ability to make selections in the two-dimensional space of fold- changes/P-values (rather than fold-change alone) can greatly enhance the selec- tivity of detection. In turn, the PFOLD algorithm has been fully integrated into our suite of analysis tools called GECKO (Gene Expression Computation and Knowledge Organisation). Turning to the analysis of entire expression profiles, as obtained for instance from time-courses, we shall present results on cross-tech- nology comparisons (e.g. Affymetrix vs. MD) and also on the biological repro- ducibility of expression profiles, including protocols for estimating false-positive rates in the presence of complex queries. We shall finally present some graphical methods for the display of expression profiles with moderate number of samples, including what we have called the “Gene Planet.” Van Belle, Patricia Gene expression of RNA for b3 integrin and CD9 in malignant melanoma and melanocytic nevi, evaluated by cDNA expression array profiling Patricia A. Van Belle, Kapaettu Satyamoorthy, Rosalie Elenitsas, Donna Strzelecki, Paulo Tahin, Hanqin Lei, M eenhard Herlyn & David E. Elder University of Pennsylvania, School of Medicine, Department of Pathology & Lab Medicine, Philadelphia, Pennsylvania 19104-4283, USA The expression pattern of mRNA was investigated in a pooled set of seven human tumourigenic vertical growth phase (VGP) melanomas and a pooled set of nine compound or dermal nevi (benign melanocytic tumours) using the Atlas Human Cancer cDNA Expression Array system. We hypothesized that expression of mRNA for beta 3 integrin would be increased in melanomas compared with nevi, because this has already been demonstrated at the protein level. We studied 587 genes from 6 functional groups, and the results were compared between the two © 1999 Nature America Inc. • http://genetics.nature.com © 1999 Nature America Inc. • http://genetics.nature.com