Connection between Genetic and Clinical Data in Bipolar Disorder

Complex diseases may be associated with combinations of changes in DNA, where the single change has little impact alone. In a previous study of patients with bipolar disorder and controls combinations of SNP genotypes were analyzed, and four large clusters of combinations were found to be significantly associated with bipolar disorder. It has now been found that these clusters may be connected to clinical data.


Introduction
Modern analytical methods, particularly in the field of molecular genetics, produce large amounts of data that pose a challenge to statistical and data-mining methods for extracting useful information [1,2]. Thus, in polygenic diseases finding disease related genetic changes, among the vast number of changes (e.g., millions of SNPs), is a daunting task. In a recent study [3], 803 SNPs in samples from 607 bipolar patients and 1355 control subjects were analyzed. All the SNPs were from genes selected based on theoretical and experimental studies that suggested that signal transduction and particular ion channels were involved in bipolar disorder [4][5][6]. The number of combinations of 3 SNP genotypes was counted in the material [3]. The theoretical number of combinations of 3 SNP genotypes taken from 803 SNPs is 2,321,319,627 (803!/3!(803 -3)!63 3) , and as many as 1,985,613,130 combinations were found in the participants. 1,719,002,329 combinations were common between controls and patients, 208,699,590 combinations were found in controls only, and 57,911,211 combinations were found in patients only, of these 45,285,770 occurred only once, and not more than 1181 combinations were shared by 9 or more patients. None of the 803 single SNPs or the nearly two billion of SNP genotype combinations showed a statistically significant association with bipolar disorder. However, among the 1181 combinations shared by 9 or more patients and no controls, four clusters of combinations, were identified that were significantly associated with bipolar disorder. Within a cluster, each patient had a personal pattern of SNP genotypes that was somewhat similar to the patterns of other patients in the same cluster, but quite different from the patterns of patients in the other three clusters, hereby suggesting an extreme degree of genetic heterogeneity [3,7].

Results
The four clusters, are shown in Table 1, 2, 3, 4. Of the 607 patients, 156 were members of the 4 clusters. The clusters contained 41, 48, 41, and 37 patients; 11 patients were members of two clusters, and no patient was a member of three clusters. The clusters contained 60, 60, 65, and 53 SNP genotypes; 29 SNP genotypes were located in two clusters, and one SNP genotype (rs1380452 positioned in the ANK3 gene) was located in three clusters.
The 156 patients were subdivided into the 4 clusters and into three groups based on three geographic areas in Scandinavia (Oslo, Aarhus, and Copenhagen). The available clinical data were not the same in the three areas. The number of hypomanic, manic and depressive episodes was available from the Norwegian patients, the number of hospital admissions and the presence of alcohol dependence was available from the patients from Copenhagen (Tables 5, 6, 7). Table 5 shows number of hypomanic and manic episodes and depressive episodes for the single patients. The median for the number of hypomanic and manic episodes is 3, and the median for the number of depressive episodes is 3.5. Patients having numbers of episodes above the median for both hypomanic and manic episodes and depressive episodes (more serious disease) were compared with patients having at least one type of episodes below or equal to the medians (p = 0.0045 for clusters 1+2+4 versus cluster 3).     Table 5. Number of hypomanic, manic and depressive episodes in patients from Oslo with bipolar disorder. 25% with more serious disease 9% with more serious disease 61% with more serious disease 15% with more serious disease Each double cell shows the number of hypomanic and manic episodes and the number of depressive episodes for a single patient. The median for the number of hypomanic and manic episodes is 3, and the median for the number of depressive episodes is 3.5. Patients having numbers of episodes above the median for both hypomanic and manic episodes and depressive episodes (bold types) are compared with patients having at least one type of episodes below or equal to the medians. p = 0.0045 for clusters 1+2+4 versus cluster 3 (Fisher's exact test, two-tailed). doi:10.1371/journal.pone.0044623.t005

Discussion
The four clusters of combinations of SNP-genotypes were statistically significantly associated with bipolar disorder; whereas biological or clinical significance of the clusters was not apparent, apart from the original selection of genes related to signal transduction [3]. These genes are shown in Table 8. The relatively little overlap between the patients in the clusters led to an analysis of available clinical data from the psychiatric departments that had recruited the patients from three locations in Scandinavia. The division of patients according to locations, clusters and availability of clinical data led to the small groups of patients shown in Tables 5-7.
Using numbers of hypomanic, manic and depressive episodes higher than the median for these episodes as an indication of severity of disease, it was found that the number of patients with more severe disease was higher in one cluster compared with the three other clusters (Table 5). This result suggests that it may be possible to connect combinations of genetic data to clinical data. The figures in Table 6 and 7 may led to similar suggestions, but although significant differences may be found between the distributions in these tables, the statistical power is low and no significant results may remain after correction for multiple testing.
Due to the relatively low number of patients as well as of clinical data, no strong conclusions can be drawn from this study. However, the results in Table 5 indicated that some genetic subgroups may be more affected by their illness than other subgroups, hereby justifying further work with combinations of genetic data as a method to connect genetic and clinical data. Hopefully, other studies with more patients, more genetic data and more clinical data will try to look at combinations of their data.

Materials and Methods
The patient sample, genes, SNP selection and genotyping, statistics and data processing regarding Table 1, 2, 3, 4 were described previously [3]. The Norwegian Scientific-Ethical Committees, the Norwegian Data Protection Agency, the Danish Scientific Committees, and the Danish Data Protection Agency approved the study. All patients gave written informed consent prior to inclusion in the project. The data in Table 5 were analyzed statistically with Fisher's exact test, two-tailed. In Tables 5, 6, 7 each box represents one patient. The numbers of hypomanic, manic, and depressive episodes in Norwegian patients were obtained by SCID [8]. The numbers of admissions in Copenhagen were obtained from patient records. Patients from Copenhagen were diagnosed with alcohol dependence when the patient was, or had been, treated in an alcohol clinic.

Author Contributions
Conceived and designed the experiments: EM PK. Analyzed the data: EM PK GLM. Contributed reagents/materials/analysis tools: HD BB IM OAA SD OM TH. Wrote the paper: EM PK.  Table 7. Alcohol dependence (1) or non-dependence (0) in patients from Copenhagen with bipolar disorder.