AlzTarget

Mendelian Randomization

Mendelian Randomization (MR) is a statistical technique used to assess causal relationships between exposures and outcomes by leveraging genetic variants as instrumental variables.

GWAS Outcome

We collected summary statistic datasets for three neurodegenerative diseases: Alzheimer's disease (AD), Parkinson's disease (PD), and amyotrophic lateral sclerosis (ALS). These datasets serve as the outcomes for the MR analysis.

Quantitative Trait Locus

We utilized three types of quantitative trait locus (QTL) datasets as the instrument variables for the MR analysis, including protein quantitative trait locus (pQTL), expression quantitative trait locus (eQTL), and splicing quantitative trait locus (sQTL). In our selection of high confidence instrumental variables, we employed four criteria:
(1) We used a false discovery rate (FDR) cutoff of less than 0.05 to identify actionable instrumental variables from all QTL datasets.
(2) The strength of the instrumental variables was evaluated using a cutoff of F-statistics greater than 10.
(3) Linkage disequilibrium (LD) pruning was performed using a reference matrix from the 1000 Genomes Project Phase 3, with an LD threshold of r² less than 0.2. This pruning was conducted using PLINK, where SNPs within a 500kb window and in LD were removed.
(4) To strictly adhere to the assumptions of MR, we excluded instrumental variables that were associated with more than two proteins or RNA expressions in all the QTL datasets. This step ensured that the instrumental variables were specific and did not have widespread effects across multiple targets.

Methods

We selected the appropriate MR methods based on the number of proposed instrumental variables (IVs) for each target. In total, we used six methods to validate all the targets:
(1) Number of IV = 1 - we used the Wald ratio estimator.
(2) Number of IV = 2 - we used the IVW fixed-effect model.
(3) Number of IV ≥ 3 - we used several MR methods to evaluate the causal effects, including the IVW random-effect model, MR-presso, Maximum Likelihood (MaxLik), Egger, and Weighted Median methods. We conducted a heterogeneity test for the IVs in the IVW and MaxLik methods. If the Cochran's Q test p-value was less than 0.05 and the I² statistics exceeded 0.5, we conducted robust IVW and MaxLik models by penalizing outlier IVs.