![[ Visit Client Website ]](images/banner.gif)
Asthma is a complex pulmonary disease with genetic and environmental components.
Our purpose with this work was to identify new genes and SNPs implicated in pathomechanism of asthma.
In this study we selected candidate genes described in the literature and from our previously completed whole genome gene expression microarray analysis of OVA induced mouse model of asthma. For further analysis we selected 308 genes. For the selection of SNPs, we created a new web-based data-mining and data-analyzing platform with the integration of public data-stores. The program performs model-based analysis on results of text and data mining processes, previous measurements and Bayesian statistical analyses. Out of the 2643 SNPs in these genes 399 SNPs were chosen and used for panel design. We designed four 48plex panels with Autoprimer software and the determination of SNPs was carried out with single base extension assays (GenomeLab SNPstream, Beckman Coulter).
Genotyping was performed in 349 asthmatic children and 461 controls. Statistical analysis was carried out by conventional frequentist analysis (chi-square test and logistic regression) and a new method developed by our team for the Bayesian analysis of relevant variables based on Bayesian model averaging (using Markov Chain Monte Carlo methods) and complex properties of Bayesian networks such as Markov Blanket sets and Markov Blanket Graphs. The developed methodology of “Bayesian, four-level, sequential analysis of relevance” is capable of incorporating diverse priors to facilitate knowledge-rich data analysis, effecting more reliable results on multidimensional genomic studies.
According to conventional analyses (logistic regression and c2 test) 36 SNPs appeared to be associated with asthma in Hungarian population while by Bayesian multilevel analysis (BMLA) only 16 SNPs came to the front and from these 8 showed association with asthma. According to our understanding a multifactorial genetic background can be revealed only by complex and aggregate method which can filter out essential and cardinal elements and alleviate the 'noise' /statistical faults. In our further studies we would like to test and confirm this conception.