High-order gene-gene interactions using STREE
Multiple interaction analyses involving various genes implicated in SLC transporters add a new dimension to assess the high-order parametric interactions within them and their impact on the overall survival of lung cancer patients. The recursive partitioning strategy was adopted to generate a survival tree in this study, demonstrating various SLC SNPs’ roles in prognosis. Figure 6a  shows the tree generated by the STREE software. The log-rank approach was used to generate the tree out of all the methods available. There were a total of four-terminal nodes found. SLC19A1G80A  was the most crucial factor as it is a root node influencing the prognosis of patients. Terminal node 5 with the highest MST was used as a reference node (13.1 months). In addition, each node’s HR is mentioned. The survival curves and the MST associated with each terminal node are illustrated inFigure 6b . It shows a difference between the median survival time of different nodes and the reference node (log-rank p 0.18), though a significant correlation was not observed. The findings of the Cox regression analysis for every terminal node are summarized inSupplementary Table 6 . Significant differences were observed in the median survival times when terminal node 4 was compared to the reference node (4.8 months versus 13.1 months; HR=2.24, 95% CI= 0.99-5.02; p=0.05). After adjusting with different covariates, including age, gender, stage, smoking, histology, regimen, ECOG, and KPS, the death risk was 1.94 times as compared to the reference node (HR=1.94; p=0.002 ). A similar trend was noted in all the other nodes; however, significant differences were not achieved.