Prioritization of Drug Targets for Neurodegenerative Diseases by Integrating Genetic and Proteomic Data From Brain and Blood

Abstract

Background: Neurodegenerative diseases are among the most prevalent and devastating neurological disorders, with few effective prevention and treatment strategies. We aimed to integrate genetic and proteomic data to prioritize drug targets for neurodegenerative diseases.

Methods: We screened human proteomes through Mendelian randomization to identify causal mediators of Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, multiple sclerosis, frontotemporal dementia, and Lewy body dementia. For instruments, we used brain and blood protein quantitative trait loci identified from one genome-wide association study with 376 participants and another with 3301 participants, respectively. Causal associations were subsequently validated by sensitivity analyses and colocalization. The safety and druggability of identified targets were also evaluated.

Results: Our analyses showed targeting BIN1, GRN, and RET levels in blood as well as ACE, ICA1L, MAP1S, SLC20A2, and TOM1L2 levels in brain might reduce Alzheimer's disease risk, while ICA1L, SLC20A2, and TOM1L2 were not recommended as prioritized drugs due to the identified potential side effects. Brain CD38, DGKQ, GPNMB, and SEC23IP were candidate targets for Parkinson's disease. Among them, GPNMB was the most promising target for Parkinson's disease with their causal relationship evidenced by studies on both brain and blood tissues. Interventions targeting FCRL3, LMAN2, and MAPK3 in blood and DHRS11, FAM120B, SHMT1, and TSFM in brain might affect multiple sclerosis risk. The risk of amyotrophic lateral sclerosis might be reduced by medications targeting DHRS11, PSMB3, SARM1, and SCFD1 in brain.

Conclusions: Our study prioritized 22 proteins as targets for neurodegenerative diseases and provided preliminary evidence for drug development. Further studies are warranted to validate these targets.

Keywords: Drug discovery; Genetics; Mendelian randomization; Neurodegenerative disease; Omics; Proteomics.

Figure 1.

Flow diagram of the study design. First, we selected independent cis-pQTLs from comprehensive pQTL datasets as genetic instruments and filtered the instruments via consistency and specificity tests. Second, we screened the human brain and blood proteomes through MR to identify candidate drug targets. Third, by applying multiple MR approaches, the causal relationships between the identified targets and diseases were further validated. Fourth, we investigated whether the protein and the disease share a common causal variant via Bayesian colocalization. Fifth, replication and correlation analyses were conducted to estimate the consistency of results within and across brain and blood. Sixth, we summarized the evidence of causality and expanded our analysis pipeline to phenome-wide to evaluate the safety of targets by predicting side effects. Finally, the druggability of the prioritized protein targets was checked. AD, Alzheimer’s disease; AGES, Age Gene/Environment Susceptibility; ALS, amyotrophic lateral sclerosis; FTD, frontotemporal dementia; KORA, Cooperative Health Research in the Region of Augsburg; LBD, Lewy body dementia; LD, linkage disequilibrium; MHC, major histocompatibility complex; MR, Mendelian randomization; MS, multiple sclerosis; PD, Parkinson’s disease; PPH4, posterior probability of hypothesis 4; pQTLs, protein quantitative trait loci; ROSMAP, Religious Orders Study/Rush Memory and Aging Project; ROSMAP-NCI, ROSMAP No Cognitive Impairment.

Figure 2.

Manhattan plots for the primary Mendelian randomization analysis of the brain and blood proteomes. By screening the human brain and blood proteomes, 19 protein-disease associations in brain (A) and 17 protein-disease associations in blood (B) were identified. Each point represents a single Mendelian randomization test ordered by chromosomal position of the sentinel cis-protein quantitative trait loci on the x-axis and −log₁₀ p value on the y-axis. The red dotted lines represent the Bonferroni multiple testing thresholds (0.05/608 for brain and 0.05/611 for blood). Proteins that survived the Bonferroni threshold are colored by their associated diseases. AD, Alzheimer’s disease; ALS, amyotrophic lateral sclerosis; BVFTD, behavioral variant frontotemporal dementia; LBD, Lewy body dementia; MS, multiple sclerosis; PD, Parkinson’s disease.

Figure 3.

Heatmap for the causal relationship assessment of the identified drug targets. The causal relationships of identified drug targets from brain (A) and blood (B) with neurodegenerative diseases were further validated by replication and multi-cis analyses; heterogeneity, pleiotropy, and directional tests; and sensitivity analysis regarding aptamer-binding effects. The depths of blue and orange represent the size of p values in Mendelian randomization analyses, and the depth of green denotes the posterior probability of colocalization. Targets with consistent evidence in all analyses were rated as a high level of causality, while targets with conflict evidence in either replication or multi-cis analysis, evidence of aptamer-binding effects, or not likely to colocalize were rated as a low level of causality and thus were not considered as credible targets. AD, Alzheimer’s disease; ALS, amyotrophic lateral sclerosis; BVFTD, behavioral variant frontotemporal dementia; LBD, Lewy body dementia; MS, multiple sclerosis; PD, Parkinson’s disease.

Figure 4.

Overall assessment of drug targets for neurodegenerative diseases. This figure illustrates 3 critical assessment aspects of drug targets: causality (x-axis), safety (y-axis), and druggability (in bold). Incredible targets are not depicted in this figure. The evidence of causality was rated according to 6 analyses and each contributes 1 point: 1) consistent in multi-cis Mendelian randomization, 2) no evidence of heterogeneity, 3) no evidence of pleiotropy, 4) true causal direction, 5) replicated in other datasets, and 6) highly likely to colocalize. The evidence of safety was approximated by the sum posterior probability of hypothesis 4 of all adverse effects that passed the Bonferroni correction. A target was rated druggable if identified in any druggable tier according to Finan’s criteria or had any related drug in the DrugBank database. Diamonds represent brain-based proteins, and discs represent blood-based proteins. Proteins are colored according to their associated diseases. Notably, the genetic association between protein GPNMB and Parkinson’s disease risk was identified in both brain and blood tissues. AD, Alzheimer’s disease; ALS, amyotrophic lateral sclerosis; MS, multiple sclerosis; PD, Parkinson’s disease.