br Statistical comparison of our study with previous

Statistical comparison of our study with previous studies (meta-analysis) was done using funnel plot and cluster analysis. Other than our study, there were 4 studies with same protocol and complete numerical information; Middleton (Spain, 2007) [13], Kin (Korea, 2014) [14], Al-Omar (Saudi Arabia, 2015) [15] and Portela (Brazil, 2017) [16]. The funnel plot was based on the Ln OR of KIR2DS5 (figure 5) while the cluster analysis was for number of haplotype B associated significant A 83 01 (figure 6). The funnel plot showed that the study of Al-Omar et al. had significantly higher impact of KIR2DS5 and Portela et al. had significantly lower impact (figure 5). For the cluster analysis 7 haplotype B related genes
ACCEPTED MANUSCRIPT
were evaluated, and among the previous studies the study of Al-Omar et al. had 5 significant association (table 4). A dissimilarity matrix was designed for these data. The most dissimilarity was between the studies Al-Omar and Portela which was 5/7 (table 5).
ACCEPTED MANUSCRIPT
4. Discussion
This study was aimed to evaluate the disease association of each KIR gene and their known HLA ligands (HLA-C and -Bw4) as well as KIR haplotypes and KIR-HLA interactions. In line of this aim we found that only KIR2DS5 was significantly associated with colorectal cancer. This association did not remain significant after applying Bonferroni's correction. No other significant single gene was found for neither KIR nor HLA. Among the haplotypes, telomeric B was significantly associated with colorectal cancer, however this association was not significant with Yate's correction. Logistic regression model showed that increasing B/X score of genotype was associated with positive trend for probability of colorectal cancer. However the low power of our study could not show this trend as a significant association.
The first limitation of our study was the low sample size of cases. This limitation resulted in low power of analysis and increasing beta error for subgroup analyses. The other limitation was that many ligands of KIR were unknown. This limitation bolds when the only significant association was for the KIR2DS5 having unknown ligand. In addition, the other limitation was lack of evaluation of other genes like APC or KRAS. Of course it needs a greater sample size. There are a lot of confounding genes associated with inflammatory bowel disease (IBD) and colorectal cancer. A whole exome sequencing study demonstrated that APC and KRAS mutation was associated with sporadic colorectal cancer in comparison to IBD-based colorectal cancer [17]. A next generation sequencing study confirmed that IBD associated colorectal cancer has different mutation [18]. However the mutations may result is progression of IBD toward colorectal cancer by itself [19].
ACCEPTED MANUSCRIPT
4.3. Interpretation
As we reported previously in a meta-analysis [3], KIR2DS5 increased risk of colorectal rectal cancer. The present study in Lur population of Iran was not excepted of course and supported the mentioned study. Although this association did not remain significant after Bonferroni's correction, but since in our study beta error was more dominant than alpha error, this correction should not be used for disproof of this significant association. From the meta-analytical point of view, based on the funnel plot (figure 5) this association was significantly more dominant in Saudi Arabia. As well the cluster analysis (figure 6) showed that in this Arabian population more number of haplotype B related genes were significantly associated with colorectal cancer [15]. In contrast to that Arabian study which had a bias for the effect of KIR2DS5 (in favor of more risk association), the study of Portela et al. showed a bias from the other side (in favor of protecting association) [16]. Our study, the study of Kim et al. [14] and the study of Middleton et al. [13] did not show bias with the pooled effect size of KIR2DS5 (figure 5). Therefore, evidence from funnel plot and cluster analysis suggests that KIR2DS5 and other haplotype B related genes have different effects in different populations. Controversially, Beksac et al. (2015) in Turkey, found that number of activating KIRs was significantly associated with long-term disease free survival [20].
4.3.2. Molecular justifications
Our rationale rational and hypothesis was that in solid tumors, sustained angiogenesis would be a hallmark. Although activation of NK cells results in further power of immune system, but on the other hand results in angiogenic cytokine release and further sustained angiogenesis. Association of haplotype B related KIR genes with other solid tumors like gastric tumors [21] and head and neck tumors [22] support this rationale.