Indiana University-Purdue University Indianapolis (IUPUI) / Cytokines mediate crucial functions in innate and adaptive immunity. They play valuable roles in immune cell growth and lineage specification, and are associated with various disease pathologies. A large number of low, medium and high throughput studies have implicated association of single nucleotide polymorphisms (SNPs) in cytokine genes with diseases. A preponderance of such experiments have not shown any causality of an identified SNP to the associated disease. Instead, they have identified statistically significant SNP-disease associations; hence, it is likely that some of these cytokine gene variants may directly or indirectly cause the disease phenotype(s). To fill this knowledge gap and derive study parameters for cytokine SNP-disease causality relationships, we have designed and developed the Disease Associated Cytokine SNP Database (DACS-DB). DACS-DB has data on 456 cytokine genes, approximately 61,000 SNPs, and 891 SNP-associated diseases. In DACS-DB, among other attributes, we present functional annotation, and heterozygosity allele frequency for the SNPs, and literature-validated SNP association for diseases. Users of the DB can run queries such as the ones to find disease-associated SNPs in a cytokine gene, and all the SNPs involved in a disease. We have developed a web front end (available at http://www.iupui.edu/~cytosnp) to disseminate this information for immunologists, biomedical researchers, and other interested biological researchers. Since there is no such comprehensive collection of disease associated cytokine SNPs, this DB will be vital to understanding the role of cytokine SNPs as markers in disease, and, more importantly, in causality to disease thus helping to identify drug targets for common inflammatory diseases. Due to the presence of rich annotations, the DACS-DB can be a good source for building a tool for the prediction of the "disease association potential (DAP)" of a given SNP. In a preliminary effort to devise such a methodology for DAP prediction, we have applied a support vector machine (SVM) to classify SNPs. Employing the SNP attributes of function class, heterozygosity value, and heterozygosity standard error, 864 SNPs were classified into two classes, "disease" and "non-disease". The SVM returned a classification of these SNPs into the disease and non-disease classes with an accuracy of 74%. By modifying various SNP and disease attributes in the training data sets, such a predictive algorithm can be extrapolated to identify potential disease associated SNPs among newly sequenced cytokine variations. In the long run, this approach can provide a means for future gene variation based therapeutic regimens.