biological pathways; it also offers a biological interpretation of diseases/traits, in particular risk phenotypes used in GWAS.(Cantor, Lange, & Sinsheimer, 2010)
Pathway based analyses have been used with several neuropsychiatric disorders. The associations between the pathway of neuronal cell adhesion molecules and either autism, schizophrenia or bipolar have been examined in two studies.(O'Dushlaine et al., 2010; K. Wang et al., 2009) In addition, a pathway analysis found ion channel activity and synaptic neurotransmission to be associated with bipolar disorder.(Askland, Read, & Moore, 2009; Holmans et al., 2009)
In this study, I used pathway analysis with one of the most common of neuropsychiatric disorders: addiction. I used the GWAS…show more content… We followed the process described in the supplementary information in Purcell’s article (Purcell et al., 2009) and divided the ALIVE GWAS sample into a female group (N=287) and a male group (N=910). We used the ALIVE GWAS male group for the discovery sample, considering four PT thresholds (arbitrary p-value thresholds below which all SNPs are summed) from the intra-ALIVE GWAS male group analyses. The ALIVE GWAS female group was used as our target sample. The male-derived polygenic risk scores were highly positively correlated with injection years in the entire ALIVE GWAS subset. The value of correlation coefficient is 0.7420077. (Supplementary Figure. 1) Using the male-derived scores in the female group with different PT thresholds, the correlation coefficient increased and then reached a plateau as the PT increased. (Supplementary Figure. 1, Figure. 7) The highest correlation coefficient is 0.1041739 when PT is 0.01. However, due to the relatively small sample size, the correlation was not statistically significant. In conclusion, the ALIVE GWAS subset suggested that injection years involve many common…show more content… Human genome references, hg19, were obtained from the UCSC Genome Browser. I selected the SNPs which reached the PT < 0.0001 and the linkage disequilibrium is under 0.5 from our genome wide association analysis results with risk phenotype as injection years. 61 clumps (interval) were formed from 78 top SNPs. I excluded 22 intervals which were not on gene regions and 8 intervals which were overlapping; 31 intervals remained in the pathway analysis. After merging and size filtering, we restricted the analysis to terms with at least 3 human genes and considered gene sets with at least 2 overlapping intervals, leaving 5,225 from the total 10,365 GO terms. A permutation procedure was conducted to rule out the probability of observing the number of intersecting intervals by chance alone. We repeated the first-pass permutations and second-pass permutations 1,000 times each to correct for bias from multiple testing. The corrected P values account for the dependence of the GO
