However, as additional microarray data will allow for a more powerful metaanalysis that reveals more common genes for both the in vivo and in vitro response, genes that do not overlap between the common in vivo and in vitro response will not always be specific for either response and merely represent the developing knowledge in this field. In the future, additional microarray data from rodents, primates, and perhaps other mammals will contribute to a further understanding of the common in vivo response and, ultimately, identification of disease mechanisms that are unique to specific agents or pathogens. We searched Y-27632 dihydrochloride Pubmed, GEO, and ArrayExpress for gene expression profiling studies related to acute lung inflammation. If corresponding microarray data were available, they were downloaded from websites indicated by the authors. Data were included in the metaanalysis if they met the following conditions: complete microarray raw or normalized data are available; a suitable uninfected or mock infection control is included in the study; time points are at most eight days after infection. Furthermore, we excluded experiments with transgenic pathogens or hosts focused on specific research questions, as these typically show inflated responses that are not representative of normal disease. Based on these ICG-001 criteria, we included 45 treatment conditions from 12 experiments. Of these studies, 4 were carried out in our laboratory and 8 were from the literature. Note that we count the data from the two related articles by Banus et al. as one study. Full details of the studies are given in Table 1. Affymetrix probe sets identifiers were converted to gene symbols using probe set annotation data downloaded from the Affymetrix website. When necessary, gene symbols in two-color or Affymetrix data files were adjusted to remove tags such as ����predicted���� and converted to uppercase symbols for further handling. All further calculations were carried out in R or Microsoft Excel. To minimize the influence of data handling procedures, we normalized all raw two-color data with the same algorithm. This consisted of a four-step approach of natural log transformation, quantile normalization of all scans, taking the sample/reference ln-ratio, and averaging replicate spot data. To remove negative values and inflated ratios, MAS4 normalized Affymetrix data were cut off at a lower value of 100, based on the findings of Grundschober et al.. MAS5 normalized Affymetrix data were used without adjustment. Affymetrix data were ln-transformed and values for replicate gene symbols were averaged. Finally, for all data sets, the average lnratio for treatment to control was calculated per gene. Treatment ratio data for the various studies were merged into one table.