Genetic insights into blood protein correlations with colorectal cancer: a Mendelian randomization study Authors: Wenjing Xu, Wei Li, Yaqiang Li, Dayu Kuai, Wei Sun, Xian Liu, Baohong Xu Published: 2024 ABSTRACT AbstractBackgroundSeveral blood proteins might be associated with the development of colorectal cancer (CRC), but many studies on this topic are often biased. By using genetic variation data, which is less influenced by environmental factors, we can better determine the causal relationship between specific blood proteins and the occurrence of colorectal cancer.MethodsData from a genome-wide association study (GWAS) on blood proteins, encompassing 1,478 proteins, and colorectal cancer (CRC) GWAS data, covering 637,693 subjects, were collected and organized. Additionally, GWAS data for obesity, diabetes mellitus (DM), and smoking were obtained for further analysis. Single nucleotide polymorphisms (SNPs) significantly associated with the exposure factors (blood proteins) were selected to ensure their independence from other confounding factors and outcomes (CRC onset), and that they only affected outcomes through blood proteins. The causal effects of Mendelian Randomization (MR) were primarily estimated using the inverse variance weighted (IVW) method, with other methods serving as supplementary approaches. The Cochran's Q-test assessed heterogeneity among SNP estimates; the MR Egger method evaluated pleiotropy; and the leave-one-out test examined the sensitivity of individual SNPs. Obesity, DM, and smoking were included in the multivariate MR analysis.ResultA total of 31 SNPs and 8 blood protein exposure factors were identified, specifically TNFRSF16 (4 SNPs), RNF8 (4 SNPs), MRM3 (4 SNPs), ST6GALNAC1 (4 SNPs), TIE1 (4 SNPs), CBP (4 SNPs), DNAJB9 (4 SNPs), and EDN2 (3 SNPs). The IVW results showed a significant causal relationship between all 8 exposure factors and colorectal cancer (P < 0.05). Among these, TNFRSF16, TIE1, and EDN2 were identified as risk factors, while the remaining five served as protective factors. The causal inference in this study was not influenced by pleiotropy or environmental factors such as obesity, diabetes, and smoking. Therefore, the results were both stable and reliable.ConclusionEight blood proteins (or genes) have been identified as having a causal relationship with the onset of colorectal cancer, suggesting their potential use as screening biomarkers and treatment targets.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12672-024-01584-y.Keywords: Blood proteins, Colorectal cancer, Causal relationship, Genetic variation, Screening biomarkers, Treatment targets Introduction Colorectal cancer (CRC) is a malignant tumor that develops in the lower digestive tract. WHO statistics show that over one million people are diagnosed with colorectal cancer worldwide each year, resulting in more than 500,000 deaths annually [1]. Its incidence rate is the third highest among global cancers, following lung and breast cancer, and it ranks fourth in mortality [2]. Thanks to the widespread use of various screening methods, the incidence rate of CRC in developed countries like the United States is decreasing. However, in some underdeveloped countries, the incidence rate is rising, particularly among younger populations, posing a serious threat to social development and public health [3, 4]. Typical clinical signs and symptoms are lacking in the early stages of this disease, making its screening challenging. As a result, most patients are diagnosed when the disease has already reached the middle to late stages [5]. Hence, the discovery of early screening markers and preventive methods is crucial. Researchers have confirmed that the development of this tumor is influenced not only by genetic factors but also by acquired and environmental factors [6]. Repeated intestinal infections can lead to inflammatory bowel disease (IBD), including ulcerative colitis and Crohn's disease. These conditions are closely linked to colorectal cancer (CRC) and are considered precancerous lesions of CRC [7]. Genetic factors can cause certain blood proteins to exhibit specific expression patterns. This was demonstrated in a study by Holm M et al. [8], in which plasma samples from 83 CRC patients were analyzed using proteomics. They found 125 proteins, including complement factor H-related 4 (CFHR4), acyl-CoA synthetase long-chain family member 5 (ACSL5), and CFHR4, to be abnormally expressed in the blood [9]. Ding D et al. analyzed the transcriptome data of CRC and identified effective blood biomarkers such as endothelial cell-specific molecule 1 (ESM1), collagen triple helix repeat containing 1 (CTHRC1), and zinc-α2-glycoprotein (AZGP1). Insulin-like growth factor-1 (IGF-1), encoded by chromosome 12, is a single-stranded peptide primarily synthesized in the liver and regulated by growth hormone. Elevated expression of IGF-1 in serum is considered a risk factor for CRC [10]. Single nucleotide polymorphisms (SNPs) in the promoter region, such as rs35767, rs2288377, and rs5742612, also regulate the expression level of this protein [11]. These SNPs represent the long-term level of IGF-1 in the body, which is more meaningful than a single serum extraction and measurement. Mendelian Randomization (MR) studies follow the Delphi genetic law of "parental alleles randomly assigned to offspring." This allows the genotype associated with a phenotype to be used as an instrumental variable to infer the association between phenotype and disease [12]. In this study, we conducted a dual-sample Mendelian randomization study using two sets of GWAS data on blood proteins and colorectal cancer. This analysis aimed to examine key SNPs and determine the impact of specific blood proteins on the risk of CRC. Materials and methods Study design This study utilized a two-sample Mendelian randomization (MR) approach, leveraging publicly available datasets to explore the influence of blood protein traits on colorectal cancer. The choice of the two-sample MR approach was based on its ability to assess causal relationships by using genetic variants as instrumental variables, which helps to mitigate confounding and reverse causation issues inherent in observational studies. Specifically, two-sample MR provides greater statistical power and flexibility when data on the exposure and outcome are derived from different, non-overlapping populations. To strengthen methodological rigor, SNPs were selected based on stringent criteria: they were genome-wide significant (p < 5 × 10−8) and not in linkage disequilibrium (LD) with other SNPs, ensuring that they independently represent the exposure of interest. This study followed the Mendelian Randomized Reporting Guidelines for Mapping the Genomic Convergence of Human Diseases, Science, 2021 [13]. Data sources Sun BB et al. conducted a gene-wide association study (GWAS) on blood proteins, analyzing data for 1478 proteins [14]. The colorectal cancer (CRC) GWAS data, which included 637,693 study subjects (14,886 CRC cases and 622,807 controls), were sourced from a large-scale multi-phenotype meta-analysis conducted by Sakaue S et al. [15]. The genome-wide association study (GWAS) data for obesity [16] originated from the FinnGen Consortium analysis, comprising 8908 obesity cases and 209,827 controls. Similarly, the diabetes GWAS data also stemmed from this analysis, encompassing 35,607 diabetes cases and 183,185 controls. The smoking GWAS data was obtained from Liu M’s GWAS analysis [17], comprising 311,629 smoking cases and 321,173 controls. Instrumental variables In this study, single nucleotide polymorphisms (SNPs) served as instrumental variables for the analysis. The screening criteria for the instrumental variables in the exposed group were as follows: (1) SNPs had to meet a significance threshold of P < 5 × 10–8 in GWAS; (2) SNPs with an r2 < 0.001 and a physical distance of more than 10,000 kb between every two genes were excluded. Subsequently, the instrumental variables in the outcome data GWAS were extracted using the filtered SNPs. The F-statistic was computed to assess the bias of weak instrumental variables [18]. An F-statistic of less than 10 indicates a weak instrumental variable, which could introduce bias to the results and should be eliminated. The specific calculation formula for the F-statistic is as follows: Among them, N represents the sample size, K represents the number of instrumental variables used, and R2 reflects the degree of exposure explained by instrumental variables. R2 = 2 × (1-MAF) × MAF × β2, where MAF is the minimum allele frequency, β is the allele effect value. MR causal effect estimation Five types of two-sample MR methods were employed to assess the causal effects of exposure on outcomes. These methods include inverse variance weighted (IVW), MR Egger, weighted median (WM), simple mode (SM), and weighted mode. Based on existing research [19], the IVW method was selected as the primary MR analysis method due to its superior performance in causal inference. The other four methods were used as supplementary analyses to validate the robustness of the results. In cases where pleiotropy was detected, the MR Egger method was applied to adjust for potential pleiotropic effects. Additionally, the same set of methods was employed to evaluate potential reverse causality. Sensitivity analysis Heterogeneity investigation The Cochran's Q test assessed the heterogeneity among SNP estimates [20], while the I2 statistic reflected the proportion of heterogeneity in the total variation of instrumental variables: I2 ≤ 0 indicated no heterogeneity, I2 = 0–25% indicated mild heterogeneity, I2 = 25–50% indicated moderate heterogeneity, and I2 > 50% indicated high heterogeneity. The specific calculation formula is as follows: Horizontal pleiotropy The MR Egger method was employed to perform a pleiotropy test on the instrumental variables. A MR Egger intercept with a P-value < 0.05 indicated significant horizontal pleiotropy of genetic variation. Leave-one-out verification The MR results for the remaining instrumental variables were calculated by sequentially removing individual SNPs. If there was a notable discrepancy between the MR effect estimation and the total effect estimation after excluding a specific instrumental variable, it suggests that the MR effect estimation is sensitive to that SNP. Statistical analysis All data calculations and statistical analyses were performed using the "TwoSampleMR" package [21] in R programming (v4.3.0). The robustness and reliability of the results were assessed through the Cochran's Q test and leave-one-out analysis. Additionally, genetic pleiotropy was examined using the MR Egger intercept method. The effect size was assessed using the odds ratio (OR) and its corresponding 95% confidence interval (CI). All statistical analyses used a two-tailed test with a significance level set at P < 0.05. Results Mendelian randomization analysis process The process flow was depicted in Fig. 1. The process flow of Mendelian randomization analysis. (1) Correlation hypothesis. (2) Independence hypothesis. (3) Exclusivity limitation. SNPs: Single Nucleotide Polymerisms. F: F statistics. MR: Mendelian randomization. Q: Cochran's Q test statistical. IVW: Inverse variance weighted. SM: Simple mode Screening of the instrumental variables According to the screening criteria for instrumental variables in this study, SNPs related to blood proteomic indicators with a significance level of P < 0.05 were included as instrumental variables, as shown in Table 1. The F-test statistic for these instrumental variables exceeds 10, suggesting that the SNPs screened in this study are predominantly strong-effect instrumental variables. Table 1Selection of instrumental variables for blood protein trait and Colorectal cancerExposureNumber of SNPsMinimum of FMedian of FMaximum of FTumor necrosis factor receptor superfamily member 16 || id: prot-a-2043457.7250045459.2288496395.62745714E3 ubiquitin-protein ligase RNF8 || id: prot-a-2573431.8708569643.36264497171.3865337rRNA methyltransferase 3, mitochondrial || id: prot-a-2575431.0023113340.94153731887.7195553Alpha-N-acetylgalactosaminide alpha-2, 6-sialyltransferase 1 || id: prot-a-2852429.8004128139.90645893124.5881533Tyrosine-protein kinase receptor Tie-1, soluble || id: prot-a-2978429.91091404112.7667361599.4549434CREB-binding protein || id: prot-a-655436.0477854541.6273689455.62419895DnaJ homolog subfamily B member 9 || id: prot-a-842430.2541117681.7841216186.7784507Endothelin-2 || id: prot-a-892340.7906637750.10978088193.672188SNPs: Single Nucleotide Polymorphisms; F: F statistics Selection of instrumental variables for blood protein trait and Colorectal cancer SNPs: Single Nucleotide Polymorphisms; F: F statistics MR causal effect estimation Five models, including MR Egger, WM, IVW, SM, and Weighted mode, were employed for the analysis of causal effects estimation. The results are shown in Fig. 2. The IVW model results indicated that Tumor Necrosis Factor Receptor Superfamily Member 16 (TNFRSF16), E3 Ubiquitin Protein Ligase (RNF8), rRNA Methyltransferase 3, Mitochondrial (MRM3), Alpha-N-Acetylgalactosamine Alpha-2,6-Sialyltransferase 1 (ST6GALNAC1), Tyrosine Protein Kinase Receptor TIE-1, Soluble (TIE1), CREB Binding Protein (CBP), DnaJ Homology Subfamily B Member 9 (DNAJB9), and Endothelin-2 (EDN2) had significant causal associations with CRC (p < 0.05). The forest plot of results of Mendelian randomization analysis. The evaluation values of the effects are presented using OR and 95% CI. The number of instrumental variables used in each model were also presented, as well as the calculated Beta values and standard errors. OR: odds ratio. CI: confidence interval Following this, we generated scatter plots (Fig. 3) to illustrate the MR analysis using different models. The results indicated significant causal associations between TNFRSF16 (Fig. 3A), RNF8 (Fig. 3B), MRM3 (Fig. 3C), ST6GALNAC1 (Fig. 3D), TIE1 (Fig. 3E), CBP (Fig. 3F), DNAJB9 (Fig. 3G), and EDN2 (Fig. 3H) with CRC. The direction of the MR analysis for different models exhibited relatively consistent slopes. Scatter plot of results of Mendelian randomization analysis. A–H Scatter plots shows the causal relationship between TNRFSF16 (A), RNF8 (B), MRM3 (C), ST6GALNAC1 (D), TIE1 (E), CBP (F), DNAJB9 (G), END2 (H) and colorectal cancer, with the slope of the straight line indicating the magnitude of the causal relationship predicted by different models Sensitivity analysis We conducted Cochran's Q test and calculated the I2 statistic to assess the heterogeneity of the significant results, as presented in Table 2. The results indicated no significant heterogeneity in the MR analysis (Cochran's Q p-value > 0.05, I2 < 50%). We employed MR Egger regression to examine the horizontal pleiotropy of instrumental variables. The statistical hypothesis test yielded P-values > 0.05, and the intercept was near 0, suggesting that horizontal pleiotropy did not affect the causal inference in this study, as depicted in Table 3. Table 2Mendelian randomization analysis heterogeneity test for the association between blood protein trait and Colorectal cancerExposureMethodQQ dfCochran's Q p-valueI2 (%)Tumor necrosis factor receptor superfamily member 16 || id: prot-a-2043MR Egger0.38006509520.8269322190Tumor necrosis factor receptor superfamily member 16 || id: prot-a-2043Inverse variance weighted0.41504336530.9371176530E3 ubiquitin-protein ligase RNF8 || id: prot-a-2573MR Egger1.23868843520.5382973280E3 ubiquitin-protein ligase RNF8 || id: prot-a-2573Inverse variance weighted1.41536982730.7019359460rRNA methyltransferase 3, mitochondrial || id: prot-a-2575MR Egger1.00813909520.604067370rRNA methyltransferase 3, mitochondrial || id: prot-a-2575Inverse variance weighted2.86681546630.4126186780Alpha-N-acetylgalactosaminide alpha-2, 6-sialyltransferase 1 || id: prot-a-2852MR Egger2.773595920.24987413427.89%Alpha-N-acetylgalactosaminide alpha-2, 6-sialyltransferase 1 || id: prot-a-2852Inverse variance weighted2.89000669530.4088967180Tyrosine-protein kinase receptor Tie-1, soluble || id: prot-a-2978MR Egger3.1899942320.20290911437.30%Tyrosine-protein kinase receptor Tie-1, soluble || id: prot-a-2978Inverse variance weighted3.21730075130.3593196816.75%CREB-binding protein || id: prot-a-655MR Egger1.7327621220.4204704560CREB-binding protein || id: prot-a-655Inverse variance weighted3.39882779830.33412280711.73%DnaJ homolog subfamily B member 9 || id: prot-a-842MR Egger1.42253472420.4910215010DnaJ homolog subfamily B member 9 || id: prot-a-842Inverse variance weighted1.57914979230.664126650Endothelin-2 || id: prot-a-892MR Egger0.5958211210.4401769150Endothelin-2 || id: prot-a-892Inverse variance weighted1.43593556220.4877424490Q: Cochran's Q test statistic; Q df: degrees of freedom for the Q test; I2 statistic reflects the proportion of heterogeneity attributed to instrumental variables in the total variability Mendelian randomization analysis heterogeneity test for the association between blood protein trait and Colorectal cancer Q: Cochran's Q test statistic; Q df: degrees of freedom for the Q test; I2 statistic reflects the proportion of heterogeneity attributed to instrumental variables in the total variability Table 3Mendelian randomization analysis of horizontal pleiotropy for the association between blood protein trait and colorectal cancerExposureMR-Egger interceptStandard errorP valueTumor necrosis factor receptor superfamily member 16 || id: prot-a-2043− 0.0434571330.232360290.868895E3 ubiquitin-protein ligase RNF8 || id: prot-a-25730.0202734230.04823160.7150964rRNA methyltransferase 3, mitochondrial || id: prot-a-25750.0435789980.031965050.3059629Alpha-N-acetylgalactosaminide alpha-2, 6-sialyltransferase 1 || id: prot-a-2852− 0.0123776830.042721780.7993001Tyrosine-protein kinase receptor Tie-1, soluble || id: prot-a-29780.0042889770.032779360.9078729CREB-binding protein || id: prot-a-655− 0.0564406810.043726650.3258665DnaJ homolog subfamily B member 9 || id: prot-a-842− 0.0337029720.085163090.7305176Endothelin-2 || id: prot-a-8920.0492883960.053774390.5276925 Mendelian randomization analysis of horizontal pleiotropy for the association between blood protein trait and colorectal cancer We performed a leave-one-out analysis and found no significant changes in the 8 proteins, demonstrating the stability of the results, as presented in Table S1. The results of the reverse causal MR analysis indicated that CRC had no causal effect on the significant blood proteomic indicators (P-value > 0.05). These results were visualized in a forest plot, as depicted in Fig. 4. Forest plot of results of reverse Mendelian randomization analysis Multivariate MR analysis We incorporated obesity, diabetes, and smoking into the analysis to perform multivariate MR analysis, aiming to assess the direct effect of significant blood proteomic indicators on colorectal cancer. In these models, Model 1, Model 2, and Model 3 represented the indirect impact of obesity, diabetes, and smoking, respectively. The results indicated that blood protein indicators continued to have a significant direct impact on CRC across these models, as outlined in Table 4. Table 4The results of multivariable Mendelian randomization analysis on the impact of blood protein trait and Colorectal cancerModelExposureNumber of SNPsOR(95%CI)P valueModel 1Tumor necrosis factor receptor superfamily member 16 || id: prot-a-204331.37 (1.33, 1.41)3.41E−108E3 ubiquitin-protein ligase RNF8 || id: prot-a-257340.85 (0.78, 0.93)0.000494746rRNA methyltransferase 3, mitochondrial || id: prot-a-257540.90 (0.86, 0.95)0.000170714Alpha-N-acetylgalactosaminide alpha-2, 6-sialyltransferase 1 || id: prot-a-285240.86 (0.76, 0.98)0.028670847DnaJ homolog subfamily B member 9 || id: prot-a-84240.85 (0.76, 0.95)0.00292256Model 2Tumor necrosis factor receptor superfamily member 16 || id: prot-a-204331.39 (1.36, 1.41)0E3 ubiquitin-protein ligase RNF8 || id: prot-a-257340.85 (0.79, 0.92)2.23E−05Tyrosine-protein kinase receptor Tie-1, soluble || id: prot-a-297831.23 (1.01, 1.50)0.043344039CREB-binding protein || id: prot-a-65540.82 (0.68, 1.00)0.045197425DnaJ homolog subfamily B member 9 || id: prot-a-84240.86 (0.78, 0.94)0.001093288Model 3Tumor necrosis factor receptor superfamily member 16 || id: prot-a-204330.84 (0.83, 0.86)4.40E−67E3 ubiquitin-protein ligase RNF8 || id: prot-a-257340.91 (0.83, 1.00)0.043775284rRNA methyltransferase 3, mitochondrial || id: prot-a-257530.85 (0.85, 0.86)0Alpha-N-acetylgalactosaminide alpha-2, 6-sialyltransferase 1 || id: prot-a-285240.79 (0.70, 0.88)3.84E−05DnaJ homolog subfamily B member 9 || id: prot-a-84231.32 (1.27, 1.37)1.68E−42Endothelin-2 || id: prot-a-89230.84 (0.83, 0.86)4.40E−67Model 1: Multivariable MR analysis of blood protein trait and obesity on Colorectal cancer; Model 2: Multivariable MR analysis of blood protein trait and diabetes mellitus on Colorectal cancer; Model 3: Multivariable MR analysis of blood protein trait and smoke on Colorectal cancer; MR: Mendelian randomization; SNPs: Single Nucleotide Polymorphisms The results of multivariable Mendelian randomization analysis on the impact of blood protein trait and Colorectal cancer Model 1: Multivariable MR analysis of blood protein trait and obesity on Colorectal cancer; Model 2: Multivariable MR analysis of blood protein trait and diabetes mellitus on Colorectal cancer; Model 3: Multivariable MR analysis of blood protein trait and smoke on Colorectal cancer; MR: Mendelian randomization; SNPs: Single Nucleotide Polymorphisms Discussion Colorectal cancer (CRC) is influenced by both genetic and acquired factors, including diet, smoking, and obesity [22]. Several proteins in the blood could be linked to the development of the disease, potentially serving as early screening biomarkers. In their study, Gao Y et al. [23] compared the diagnostic efficacy of various biomarkers in the serum of CRC patients. They observed that the sensitivity ranking was CEA > CA72-4 > CA19-9 > ferritin > CA125. In their study, Müller D et al. [24] employed gene transcriptomics and confirmed that several biomarkers, including SDC2, VIM, APC, MGMT, SFRP1, SFRP2, and NDRG4, might be associated with the progression of CRC, particularly in terms of DNA methylation. Nonetheless, the methodologies, tools, and samples used in these surveys were all subject to bias and confounding factors, rendering it challenging to accurately investigate the causal relationship between blood proteins and the onset of colorectal cancer. Genes are inherited at birth and are less influenced by environmental and behavioral factors. Genetic variation data associated with environmental exposure can be used to understand the effects of specific genetic phenotypes on diseases [25]. In our study, we included data from 1478 protein GWAS and 637,693 CRC study subjects. After excluding SNPs that did not meet certain criteria, we identified 31 SNPs for 8 blood proteins. Using 5 models, we analyzed the SNP effects on the association between these proteins and CRC incidence. The IVW model showed significant effects for all 8 proteins (P < 0.05), while the other models did not find significant causal associations. Considering IVW's consideration of heterogeneity, we deemed its results more reliable. Among these proteins, TNFRSF16, TIE1, and EDN2 increased the risk of CRC, while the other 5 proteins reduced this risk and were considered protective. The TNFRSF16 gene encodes a protein that belongs to the TNF receptor family, characterized by cysteine-enriched regions, death domains, and TNF receptor-associated factor (TRAF) binding sites [26]. Functionally, TNFRSF16, also known as the p75 neurotrophin receptor (NTR), is a member of the nerve growth factor receptor family. Its binding to neurotrophins can activate the NF-kB pathway, promote inflammation, and induce cell apoptosis, proliferation, and differentiation [27]. Demir IE et al. [28] noted in their study that nerve growth factor and its receptors TrkA and p75 (NTR) are overexpressed in CRC. Targeting TrkA/NGF through selective small molecule inhibitors or antibodies shows significant potential in CRC treatment, providing a promising therapeutic direction that aligns with the findings of our study. The TIE1 gene encodes a receptor tyrosine kinase protein with immunoglobulin-like and EGF-like domains, which belongs to the tyrosine protein kinase family [29]. Nakayama T et al. [30] found that TIE1 and its ligands are positively expressed in colorectal cancer cells, and these receptor-ligand complexes contribute to the differentiation and progression of human colorectal cancer cells. This supports the therapeutic potential of targeting TIE1 to inhibit cancer progression, consistent with our study findings. ET-2, encoded by the EDN2 gene, is produced in human vascular endothelial cells and primarily functions by binding to two receptors, EDNRA and EDNRB. Wu F et al. [31] constructed a network of TF-gene-miRNAs in CRC and identified genes such as GCNT4, EDN2, MYC, and WT1 as hub genes closely interacting with other RNAs in the network. Importantly, our study is the first to identify ET-2 as a protein directly implicated in promoting colorectal cancer development. The identification of these proteins as potential biomarkers and therapeutic targets highlights their clinical relevance, suggesting that future research should focus on developing targeted therapies and validating these findings in clinical settings to enhance CRC diagnosis and treatment strategies. The other five blood proteins can reduce the incidence of CRC. E3 ubiquitin protein ligase, encoded by the RNF8 gene, is involved in DNA damage repair. However, Ren L et al. [32] suggested in their study that RNF8 could induce c-Myc expression mediated by β-catenin, promoting colon cancer proliferation. In Xu Y et al.'s study [33], RNF8 was found to be overexpressed in cancer, positively correlating with p-Akt expression and low survival rates in non-small cell lung cancer patients. These findings present conflicting results, possibly due to RNF8's involvement in various cancer pathways and regulations. However, our research, based on large-sample analysis, provides more reliable results. The protein encoded by ST6GALNAC1 is a glycosyltransferase that plays a role in the biosynthesis of O-glycans and glycan structure biosynthesis [34]. This enzyme's influence on the development of colorectal cancer has been identified for the first time. CREB binding protein (CBP) is a transcriptional coactivator involved in various cellular processes, including DNA repair, cell growth, differentiation, and apoptosis. Du C et al. [35] highlighted in their study that individuals with low CBP expression in CRC tissue exhibit a lower long-term disease-free survival rate and increased early progression. Additionally, 5-Fluorouracil targets this protein for the treatment of colorectal cancer. The protein encoded by DNAJB9 belongs to the J protein family, which is known for its role in protecting stressed cells from apoptosis [36]. However, this study is the first to discover the protective effect of this gene on colorectal cancer. We performed a sensitivity analysis on the study results and found no influence from horizontal pleiotropy. Exclusion of certain variables did not significantly alter the outcomes, and reverse MR analysis indicated no substantial impact of colorectal cancer on the blood proteome. Additionally, the introduction of external factors such as obesity, diabetes, and smoking did not significantly affect the study results. Hence, the stability and reliability of our findings are affirmed. Limitations This study has several limitations. Firstly, only the IVW model yielded significant results among the five model algorithms for SNP synthesis, suggesting potential heterogeneity in the results. Secondly, the number of SNPs for each blood protein was limited, typically ranging from 3 to 4, which is relatively small. Thirdly, the results obtained in this study were solely based on data analysis, and their validity should be confirmed through both in vivo and in vitro experiments. Conclusions The Mendelian randomization findings in this study indicated a causal link between 8 blood proteins and colorectal cancer development. Notably, among these, only three proteins (TNFRSF16, TIE1, and EDN2) were associated with an increased risk of colorectal cancer, suggesting their potential as risk biomarkers. However, it is important to critically consider the implications of these limited findings. The fact that only three proteins were linked to increased CRC risk suggests that the role of blood proteins as biomarkers may be more selective and context-dependent than initially anticipated. This highlights the need for further research to validate their clinical utility, both in terms of diagnostic specificity and their potential as therapeutic targets. Moreover, additional experimental validation is required to confirm these associations and fully understand the underlying biological mechanisms. Supplementary Information Additional file 1. Acknowledgements None. Author contributions BX designed the study. WX, WL, YL, DK, WS, and XL performed data analysis. WX drafted the manuscript. BX revised the manuscript. All authors read and approved the final manuscript. Funding None. Availability of data and materials The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author. Declarations Ethics approval and consent to participate Not applicable. Informed consent Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Additional file 1. Data Availability Statement The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.