Table 2.
Significant associations between pre-treatment DNAR gene expression and pathological response, disease free survival or overall survival.
| Gene | LogFC | P-value | |
|---|---|---|---|
| Higher expression associated with lack of pathological response | |||
| MPG | −4.035 | 0.010 | |
| MEN1 | −3.024 | 0.010 | |
| USP3 | −2.880 | 0.020 | |
| ALKBH1 | −3.288 | 0.025 | |
| MGMT | −1.383 | 0.028 | |
| PARP3 | −2.543 | 0.035 | |
| KIN | −2.154 | 0.041 | |
| Higher expression associated with DFS <1 year | |||
| CHAF1B | −0.648 | 0.004 | |
| BCCIP | −0.704 | 0.005 | |
| RAD51L1 | −0.597 | 0.014 | |
| ESCO1 | −0.639 | 0.015 | |
| RDM1 | −0.689 | 0.020 | |
| MEN1 | −0.599 | 0.022 | |
| ERCC1 | −0.624 | 0.025 | |
| H2AFX | −0.503 | 0.027 | |
| NEIL3 | −0.581 | 0.027 | |
| EME1 | −0.573 | 0.037 | |
| ALKBH1 | −0.421 | 0.041 | |
| RAD51 | −0.561 | 0.046 | |
| GTF2H4 | −0.359 | 0.046 | |
| Gene | Hazard ratio (With 95% CI) | P -value | |
| Higher expression associated with worse OS | |||
| ERCC1 | 3.113 | (1.56–6.22) | 0.001 |
| ERCC6 | 10.505 | (1.71–64.55) | 0.011 |
| HINFP | 13.008 | (1.52–111.2) | 0.019 |
| RAD51L1 | 3.611 | (1.23–10.58) | 0.019 |
| HMGB1 | 4.381 | (1.26–15.26) | 0.020 |
| ALKBH1 | 5.174 | (1.28–20.98) | 0.021 |
| FSBP | 3.907 | (1.18–8.11) | 0.021 |
| NEIL1 | 4.161 | (1.19–14.52 | 0.025 |
| GTF2H2 | 3.865 | (1.16–12.84) | 0.027 |
| ESCO2 | 2.796 | (1.05–7.42) | 0.039 |
| BCCIP | 2.634 | (1.01–6.90) | 0.049 |
Principal component analysis was carried out to ensure that tumor and non-malignant epithelium showed segregation and to check for any technical artifacts. Differential expression (log fold change) of DNAR genes in tumor tissue from 38 patients with esophageal adenocarcinoma was compared between pathological non-responders and pathological responders, and between patients with DFS <1 year and DFS >1 year, using linear regression. Pre-treatment gene expression levels significantly associated with OS were identified by Cox proportional hazards regression. Unadjusted P values are presented to enable identification of all hits appropriate for hypothesis testing and subsequent validation, albeit with recognized risk that some of the hits might appear by chance.