This paper introduces a new method for Mendelian randomization (MR) analysis using summary data, called the mode-based estimate (MBE). The MBE is designed to provide a robust estimate of the causal effect of exposures on outcomes, even in the presence of horizontal pleiotropy, which occurs when genetic variants affect the outcome through mechanisms other than the exposure of interest. The method relies on the ZEro Modal Pleiotropy Assumption (ZEMPA), which posits that the most common causal effect estimate among genetic instruments is a consistent estimate of the true causal effect, even if most instruments are invalid. The MBE is evaluated through simulations that mimic the two-sample summary data setting and is applied to investigate the causal effect of plasma lipid fractions and urate levels on coronary heart disease risk. The results show that the MBE has less bias and lower type-I error rates than other methods under the null hypothesis in many situations. While its power to detect a causal effect is smaller compared to the inverse variance weighting (IVW) and weighted median methods, it is larger than that of MR-Egger regression. The MBE requires smaller sample sizes than those typically available from genome-wide association study (GWAS) consortia. The MBE is less sensitive to violations of the instrumental variable assumptions compared to other methods, and it should be used in combination with other approaches in sensitivity analyses. The method is implemented by using the mode of the smoothed empirical density function of all ratio estimates of the causal effect. This approach allows for different weights to be assigned to different instruments and is straightforward to implement. The paper also discusses the performance of the MBE under various simulation scenarios, including the causal null hypothesis and the presence of horizontal pleiotropy. The results show that the MBE is less biased than IVW and weighted median methods when the InSIDE assumption is violated, but it is less biased than MR-Egger regression in cases of large proportions of invalid instruments. The MBE is also compared to other MR methods, including the weighted median and IVW, in terms of power and precision. The study concludes that the MBE is a robust method for causal inference in observational studies using summary data, and it should be used in combination with other approaches to assess the robustness of results against violations of the instrumental variable assumptions. The MBE is particularly useful in the two-sample setting and when using precise summary association results. The method is also compared to other approaches, such as the weighted median and IVW, in terms of performance and robustness. The paper highlights the importance of considering the assumptions underlying different MR methods and the need for further research to improve the accuracy and reliability of causal inference in observational studies.This paper introduces a new method for Mendelian randomization (MR) analysis using summary data, called the mode-based estimate (MBE). The MBE is designed to provide a robust estimate of the causal effect of exposures on outcomes, even in the presence of horizontal pleiotropy, which occurs when genetic variants affect the outcome through mechanisms other than the exposure of interest. The method relies on the ZEro Modal Pleiotropy Assumption (ZEMPA), which posits that the most common causal effect estimate among genetic instruments is a consistent estimate of the true causal effect, even if most instruments are invalid. The MBE is evaluated through simulations that mimic the two-sample summary data setting and is applied to investigate the causal effect of plasma lipid fractions and urate levels on coronary heart disease risk. The results show that the MBE has less bias and lower type-I error rates than other methods under the null hypothesis in many situations. While its power to detect a causal effect is smaller compared to the inverse variance weighting (IVW) and weighted median methods, it is larger than that of MR-Egger regression. The MBE requires smaller sample sizes than those typically available from genome-wide association study (GWAS) consortia. The MBE is less sensitive to violations of the instrumental variable assumptions compared to other methods, and it should be used in combination with other approaches in sensitivity analyses. The method is implemented by using the mode of the smoothed empirical density function of all ratio estimates of the causal effect. This approach allows for different weights to be assigned to different instruments and is straightforward to implement. The paper also discusses the performance of the MBE under various simulation scenarios, including the causal null hypothesis and the presence of horizontal pleiotropy. The results show that the MBE is less biased than IVW and weighted median methods when the InSIDE assumption is violated, but it is less biased than MR-Egger regression in cases of large proportions of invalid instruments. The MBE is also compared to other MR methods, including the weighted median and IVW, in terms of power and precision. The study concludes that the MBE is a robust method for causal inference in observational studies using summary data, and it should be used in combination with other approaches to assess the robustness of results against violations of the instrumental variable assumptions. The MBE is particularly useful in the two-sample setting and when using precise summary association results. The method is also compared to other approaches, such as the weighted median and IVW, in terms of performance and robustness. The paper highlights the importance of considering the assumptions underlying different MR methods and the need for further research to improve the accuracy and reliability of causal inference in observational studies.