Uncovering gene networks in cancer by computational systems biology

The aim of the study was to develop the data analysis strategy for discovery of genetic background of radiosensitivity in GWAS type of study. Identification of polymorphisms and genes responsible for organism’s radiosensitivity may allow for better understanding the process of carcinogenesis and impact of radiotherapy. Materials and methods: The population under investigation is composed of fraternal twins, monozygotic twins and non-related persons, and the analysed sample consists of 130 randomly chosen individuals in total (66, 30 and 34 respectively). Two type of data were collected, first was the result of genotyping of 567 096 polymorphisms, the second one includes 9 gene qPCR expressions measured after the irradiation of 2Gy and in normal conditions, accompanied by activity of γH2AX test on radiosensitivity. Standardized fold changes per each gene were calculated. The first step of the analysis was focused on the subpopulation of non-related persons. To each of SNP, basic statistics were calculated. Also three different models of SNP-gene expression interaction (genotype, recessive and dominant) were investigated and the criterion of minimum p-value was used to choose the most suitable one per each SNPgene pair. From the characteristic p-value to each SNP only those, which show statistic significance were taken to further analysis. The validation of the obtained set of candidate markers was done with the use of fraternal twins, allowing for the discrimination some of the SNPs not directly related to the radiosensitivity. The final verification was performed on monozygotic twins. Results: During the first step, the analysis of interaction between every SNP and radiosensitivity test and gene expression results was performed among the non-related persons. The hypothesis on normality of gene expression distribution within each subpopulation was verified by Lilliefors test, and after the correction on multiple testing, the hypothesis could not be rejected at 0.05 significance level. ANOVA algorithm was applied to investigate the genotype model of interaction and t-test to dominant and recessive ones. Having a set of three of p-values obtained for each model, the best one with minimum p-value was chosen. The exemplary distribution of the final models of genetic interaction between SNPs and γH2AX look as follow: to genotype model belong 1 868 SNPs (FDR 7 371), to dominant model belong 203 875 SNPs (FDR 13 950), to genotype model belong 243 247 SNPs (FDR 15 910) and total number is 448 990 SNPs (FDR 22 450). One can notice that only 0.41 percent of results are the genotype type of interactions, the most frequent are dominant and recessive models. The final set of 55 289 (p0.05), where FDR is 22 450, candidate radiosensitivity markers was validated with the use of heterozygotic twins and above 95% were filtered out. The final step of marker verificaion among monozygotic twins is going to be performed soon. Conclusions: Preliminary analysis shows that different models of interaction must be included in the investigation of the genetic background of biological phenomena. 116 Session 6. Molecular Mechanism of Cancer Development and Progression 48th Congress of the Polish Biochemical Society Looking at allelic frequencies and genotyping results only would not allow for finding the candidate biomarkers.