Figure 1.
Enrollment, attrition, drug treatments, and outcomes of patients in all samples.
Table 1.
Characteristics of patients in derivation sample, validation sample, and cross-validation sample.
Table 2.
The SNPs most strongly associated with SSRI response (P <0.05 after FDR correction) and the strongly associated polymorphisms in SLC6A4 from our previous study [9].
Figure 2.
Linkage disequilibrium (LD) and haplotype structure of (a) TPH2, (b) GRIK2, (c) SLC6A4, and (d) GAD1.
The LD structure in the lower panel is based on the measure of r2. Dark red indicates strong LD between two markers with high r2 and a logarithm of odds (LOD) score of greater than 2.0. Haplotype frequencies of responders and nonresponders are also shown in each box in order (responders: nonresponders). The figure was prepared with LocusView2.0 (http://www.broad.mit.edu/mpg/locusview). Significant SNPs inscribed in red are plotted with their association analysis p values (as –log10 values) in the upper panel. Haplotype blocks and estimated haplotypes for each block are presented in the middle panel. (a) Among 30 SNPs screened in TPH2, four SNPs (colored red), rs17110532, rs4760815, rs11179027 and rs17110747, were significantly associated (see Table 2). The third (H3), fourth (H4) and fifth haplotype (H5) blocks were significantly associated with drug response (P<0.05 after FDR correction). (b) Among 78 SNPs in GRIK2, two SNPs (colored red), rs543196 and rs572487, were significantly associated. Four SNPs, rs580543, rs6940676, rs513216, and rs673318, adjacent to the peak SNPs also showed high associations. The eighth (H8) and ninth (H9) blocks were significantly associated with response (P<0.05 after FDR correction). (c) Among 12 SNPs in SLC6A4, two SNPs (colored red), rs2066713 and rs2020942, were significantly associated with response. The first haplotype (H1) block was significantly associated with response (P<0.05 after FDR correction). (d) Among ten SNPs in GAD1, two SNPs (colored red), rs3828275 and rs12185691, were significantly associated (see Table 2). No haplotype blocks were significantly associated with response.
Table 3.
Genotypic combinations of haplotype-SNP (HAP-SNP) prediction model.
Figure 3.
Clinical performance of selective serotonin reuptake inhibitor (SSRI) response prediction model using genetic information.
Results of genetic prediction of response or nonresponse to SSRIs (a) in initial derivation sample (n = 239) and (b) validation sample (n = 176) of completer patients with major depression.