The IVs needed for GM, IC, and TC studies were screened using the aforementioned method. The number of SNPs in GM, IC, and TC we studied ranged from 6 to 22, 12 to 31, and 22, respectively. Additionally, the F-statistic of SNPs screened in this study was at least 15. These data are detailed in Supplementary Material Table V1-V20.

After MR analysis, GM species with a causality (P < 0.05) with TC were screened among 473 GM. Ultimately, 9 GM species were found to have a causal association with TC, with no reverse causality identified (Fig. 2). Using IVW as the primary method, specific GM species demonstrated notable odds ratios (OR), indicating their strength of association with TC. For example, s_Veillonella rogosae had an OR of 1.1871, which implies that an increase in its abundance is associated with an approximately 18.7% higher risk of TC. Similarly, s_Bacteroides stercoris had an OR of 1.6441, suggesting a 64.4% increased risk of TC, highlighting its strong potential role in tumorigenesis. The highest effect was seen with f_Mycobacteriaceae (OR = 2.1736), indicating that its presence could more than double the risk of TC (a 117.4% increase), pointing to a particularly significant influence.

MR analysis OR and P value of 9 gut microbiota on thyroid cancer

These effect sizes illustrate the varying degree of impact that different GM species may have on TC. The OR values reflect the potential risk modulation of TC as mediated by GM, providing insight into which species might be prioritized for further research. This highlights the importance of specific GM species not just in terms of their presence but also their quantitative effect on TC risk. The summary results of the MR analysis, including detailed ORs for all GM species, are provided in Fig. 3 and the Supplementary Materials (Figures S1-9, Tables T1-9, U1-9, and V1-9).

MR analysis P value of 9 gut microbiota on thyroid cancer

After conducting MR analysis, we screened 731 IC to identify those causally related to TC based on GM (P < 0.05), ultimately identifying 10 types that were causally related to TC (Fig. 4). When IVW serves as the primary method to evaluate the causality of IC on TC in the MR analysis, Switched Memory AC (OR = 1.075), CD25 on CD39 + CD4 + (OR = 1.033), IgD- CD38dim AC (OR = 1.0876), Effector Memory Double Negative (EMDN) (CD4-CD8-) AC (OR = 1.105), B cell %lymphocyte (OR = 1.0537), CD28 + DN (CD4-CD8-) AC (OR = 1.1009), CCR2 on CD14 + CD16 + monocyte (OR = 1.024), SSC-A on granulocyte, the expression of CD11c on myeloid Dendritic Cells (DC) (OR = 1.0568), and CD11b on Mo MDSC (OR = 1.0508) were correlated with TC. The detailed results of all MR analyses are shown in Fig. 5 and Supplementary Figures (S21-30, T21-30, U21-30, and V10-19).

MR analysis P value of 10 immune cell on thyroid cancer

MR analysis OR and P value of 10 immune cell on thyroid cancer

When IVW serves as the primary method to evaluate the causality of GM on IC in the MR analysis, s_Veillonella rogosae on Switched Memory AC (OR = 1.2306), s_Veillonella rogosae on IgD- CD38dim AC (OR = 1.2288), s_Massiliomicrobiota on B cell % lymphocyte (OR = 1.1992), s_Massiliomicrobiota on SSC-A on granulocyte (OR = 1.2347), s_Phocea massiliensis on CCR2 on CD14 + CD16 + monocyte (OR = 1.3946), and g_Saccharomonospora on EM DN (CD4-CD8-) AC (OR = 1.5444). The above results revealed that GM had a positive correlation with IC. Detailed results of all MR analyses are shown in Fig. 6 and Supplementary Figures (S10-20, T10-20, U10-20, and V20).

MR analysis OR and P value of 9 gut microbiota on 10 immune cells

MR analysis revealed that 10 IC mediated the causality between 9 GM and TC. Notably, 'EM DN (CD4-CD8-) AC' exhibited the highest mediation ratio (beta_P of 15.9%) in the pathway between g_Saccharomonospora and TC, suggesting that this immune cell type may play a particularly significant role in influencing tumor progression. The strong mediation by 'Switched Memory AC' (beta_P of 8.7%) and 'IgD-CD38dim AC' (beta_P of 10.1%) in the relationship between s_Veillonella rogosae and TC highlights the potential importance of adaptive immune memory responses in modulating cancer outcomes. These cell types are known for their roles in sustaining long-term immune surveillance and response, which could explain their notable mediation effects.

In contrast, 'CCR2 on CD14 + CD16 + monocyte' had a lower mediation ratio (beta_P of 2.2%) in the pathway between s_Phocea massiliensis and TC, possibly reflecting a more specialized or context-dependent role of this monocyte subtype in the immune response to tumorigenesis. The involvement of 'CD28 + DN (CD4-CD8-) AC' (beta_P of 9.7%) in mediating the effect of s_Bacteroides stercoris on TC may indicate a crucial role in immune regulation, given the dual functions of CD28 in co-stimulatory signaling for T cells.

The varying mediation ratios observed suggest that some IC, such as 'EM DN (CD4-CD8-) AC' and 'Switched Memory AC,' may have broader and more impactful roles in the interplay between GM and TC, potentially due to their involvement in maintaining immune homeostasis and response modulation. Meanwhile, mediators like 'CD11b on Mo MDSC' (beta_P of 7.0%) and 'CD11c on myeloid DC' (beta_P of 5.6%) underscore the influence of myeloid lineage cells in immune suppression and tumor immune escape. These differences emphasize the complexity of the immune response and the varied influence of different IC in mediating GM's effects on TC, warranting further experimental validation to uncover the mechanisms driving these interactions. These data are detailed in Table 2.

This MR study found no evidence to support the existence of horizontal pleiotropy, as indicated by the MR-Egger and MR-PRESSO. Leave-one-out study revealed that independent SNPs didn’t significantly influence the overall effect in the MR analyses of GM-TC and TC-GM (Supplementary Material Figure U1-U30). Overall, the findings of the MR study were confirmed to be robust through the conducted sensitivity analyses.