MR analysis

To determine the causal association between serum chemokines and cancers, a two-sample MR Analysis was performed. A total of three common MR Analysis methods have been used, including inverse-variance weighted (IVW) (Huang et al., 2022), MR-Egger regression (Wu et al., 2020), weighted median (Li et al., 2022), weighted mode and simple mode methods are supplemented. According to the survey, the IVW test exhibits superior advantages compared to additional methods (Lin et al., 2021). And it has been used as the primary MR Analysis method in most studies (Yang M. et al., 2023; Yang Y. et al., 2023; Ding et al., 2023; Li et al., 2023). Similarly, IVW was used as the main test method in our study, while other methods were used as references. In addition, the MR-Egger regression test and MR-Presso were used to verify the existence of horizontal pleiotropy, and p-value < 0.05 is considered to be horizontal pleiotropy. To ensure the validity of our findings, we conducted leave-one-out sensitivity analysis to ascertain whether a single SNP is responsible for driving the results. Based on the causal relationship between 38 chemokines and cancer, the more conservative Bonferroni method was used to correct for significance results. Before correction, p < 0.05 was a significant result, and after correction, p < 0.0013 was a significant result. Results with p < 0.05 but higher than 0.0013 were considered for potential causal associations (Sedgwick, 2014; Larsson et al., 2017). All statistical tests were performed in two-sample MR and MR-PRESSO packages. Moreover, Heterogeneity test results were significant (p < 0.05), which was considered to be heterogeneity among IVs.

To investigate the bidirectional causal relationship between cancer and chemokines, we performed a bi-directional Mendelian randomization (bi-MR) analysis. Cancers were used as exposure variable and chemokines as outcome variable.