Table 2.
The best set of 17 SNPs selected by the varSelRF from 57 SNPs belonging to the DNA repair pathway; the San Francisco Bay Area Adult Glioma Study
| Gene symbol | SNPs | Alleles* | No. cases HW/HT/HV | No. controls HW/HT/HV | Ranking in the original RF† | Frequency selected as part of the best SNPs in 200 bootstrap samples | Log-additive model OR (95% CI)‡ | Ranking by P value from logistic regression |
|---|---|---|---|---|---|---|---|---|
| EXO1 | rs1047840 | A/G | 55/42/15 | 29/59/22 | 1 | 0.99 | 0.54 (0.37–0.80) | 1 |
| CHAF1A | rs9352 | C/T | 20/57/35 | 32/48/32 | 15 | 0.78 | 1.29 (0.90–1.87) | 10 |
| EME1 | rs12450550 | C/T | 51/52/9 | 68/35/9 | 3 | 0.71 | 1.46 (0.96–2.21) | 3 |
| BRCA1 | rs799917 | C/T | 51/51/10 | 55/38/19 | 2 | 0.59 | 0.90 (0.62–1.32) | 34 |
| EXO1 | rs1776148 | A/G | 44/57/11 | 46/47/18 | 10 | 0.55 | 0.90 (0.61–1.33) | 35 |
| BRCA1 | rs16941 | A/G | 51/48/11 | 55/36/18 | 6 | 0.47 | 0.93 (0.64–1.37) | 45 |
| ERCC6 | rs2228526 | A/G | 63/43/4 | 70/28/6 | 7 | 0.38 | 1.26 (0.79–2.01) | 9 |
| ATM | rs1801516 | A/G | 92/18/2 | 76/31/5 | 5 | 0.34 | 0.51 (0.30–0.88) | 2 |
| ERCC6 | rs2228527 | A/G | 60/48/4 | 72/33/6 | 14 | 0.29 | 1.32 (0.84–2.09) | 15 |
| CCNH | rs2266690 | C/T | 71/34/7 | 63/45/4 | 17 | 0.27 | 0.87 (0.56–1.37) | 31 |
| MSH4 | rs5745325 | A/G | 63/38/11 | 62/43/6 | 12 | 0.25 | 1.11 (0.73–1.68) | 37 |
| BRCA1 | rs1799966 | A/G | 53/49/9 | 56/39/17 | 4 | 0.24 | 0.89 (0.60–1.31) | 32 |
| POLE | rs5744934 | A/G | 71/36/5 | 81/28/3 | 9 | 0.24 | 1.43 (0.88–2.33) | 6 |
| ERCC6 | rs4253211 | C/G | 82/18/0 | 77/26/0 | 13 | 0.19 | 0.62 (0.31–1.23) | 9 |
| POLQ | rs3218651 | A/G | 72/24/2 | 75/20/2 | 11 | 0.16 | 1.16 (0.65–2.09) | 36 |
| POLQ | rs3218634 | C/G | 94/15/2 | 101/11/0 | 8 | 0.005 | 1.75 (0.82–3.71) | 5 |
| POLQ | rs532411 | C/T | 95/15/2 | 101/11/0 | 16 | 0.00 | 1.73 (0.82–3.67) | 7 |
| Bootstrap (.632+) estimate of prediction error: 45.33% | ||||||||
NOTE: The best set of SNPs was determined using an iterative process of fitting RF and dropping the lowest ranked SNP. The smallest set of SNPs with the lowest OOB error rate was considered the best set.
Abbreviations: HW, homozygous wild type; HT, heterozygous; HV, homozygous variant. OR, odds ratio; 95% CI, 95% confidence interval.
*
Bold letters indicate minor alleles.
†
The RF built with all SNPs in the DNA repair pathway.
‡
Adjusted for sex and age using unconditional logistic regression, with the major allele as the reference group.