250420 몇 가지 Mendelian randomization 논문들

몇 가지 mendelian randomization 논문을 보고 다음과 같은 분석들을 추가로 하면 좋을 것 같다고 생각함:  

1. Mendelian randomization 결과의 orthogonal evidence를 위해 SMR + HEIDI, TWAS, PWAS 등의 분석을 추가로 한다

2. Protein을 target하는 알려진 drug가 아직 없는 경우 moldecular structure를 이용해 drug candidate을 찾는다. 

 

1.

Jiang Y, Wang Y, Guo J, et al. Exploring potential therapeutic targets for asthma: a proteome-wide Mendelian randomization analysis. J Transl Med. 2024;22(1):978. Published 2024 Oct 29. doi:10.1186/s12967-024-05782-8

 

 

 

 

 

Furthermore, SMR analysis was used as a complementary method to explore the causal relationship between proteins and asthma. Unlike traditional MR, the SMR and HEIDI methods utilize summary-level data from GWAS and pQTLs studies to test whether the protein and phenotype are correlated due to shared causal variants [23]. To combine the findings from MR and illustrate the causality of the results, positive results from either two-sample MR or SMR were considered indicative of causal effects. SMR software (SMR v1.3.1) was used for SMR and HEIDI tests.

 

Several databases, including DrugBank [29], PubChem [30], Therapeutic Target Database [31], and ChEMBL [32], were queried to assess the current status of existing therapies targeting the identified proteins with co-localization evidence

The DSigDB database was used in this study to predict potentially effective intervention drugs for the proteins PCDH12, LTB, and MAX, for which no targeted drugs were found. The top 10 potential chemical compounds were identified based on adjusted P values, as shown in Table S8. Due to the unresolved macromolecular structure of LTB, Autodock was employed to determine the binding sites and interactions between the 6 drug candidates and the proteins encoded by MAX. Subsequently, we generated the corresponding binding energies for each interaction, as detailed in Table S9 and Fig. 3B-G. Notably, the MAX-digitoxigenin interaction exhibited the lowest binding energy of − 8.24 kcal/mol, reflecting a highly stable affinity.To verify the stability of MAX and its predicted small molecule drugs, we docked classical asthma targets along with their corresponding small molecule drugs and calculated their binding energies. For instance, formoterol binds to the beta-2 adrenergic receptor with a binding energy of − 5.54 kcal/mol, while hydrocortisone interacts with annexin A1 at − 6.54 kcal/mol and the glucocorticoid receptor, demonstrating a strong affinity of − 11.94 kcal/mol, among others. These results indicate similar strong binding affinities, consistent with the binding energies for other medications predicted by MAX (ranging from − 4.2 kcal/mol to − 8.24 kcal/mol). Detailed results can be found in Table S10 and Supplementary Figure.

 

2. 

Xie Z, Feng Y, He Y, Lin Y, Wang X. Identification of potential drug targets for pelvic organ prolapse using a proteome-wide Mendelian randomization approach. Sci Rep. 2025;15(1):8291. Published 2025 Mar 10. doi:10.1038/s41598-025-92800-4

Drug candidate prediction

We used the DSigDB to predict potential therapeutic drug candidates. Using a significance cut-off of P < 0.05, we identified drugs that target both EFEMP1 and MFAP4 and considered their toxicity profiles. Three promising drug candidates were selected (Supplementary Table 13): dasatinib (CTD 00004330), progesterone (CTD 00006624), and retinoic acid (CTD 00006918).

Molecular docking

We performed molecular docking studies to assess the interactions between potential drugs and target proteins, focusing on binding affinities and energy calculations (Figs. 6A–F). Lower binding energy values indicate more stable binding. Our studies revealed high binding affinities between the drug candidates and the two protein targets, with docking energies below − 5 kJ/mol, suggesting strong and stable binding and indicating that these drugs can effectively target genes (Supplementary Table 14).

 

Drug candidate prediction and molecular docking

Evaluating protein-drug interactions is critical for determining whether a gene can serve as a viable drug target. We used the Drug Signatures Database (DSigDB)34 to predict potential drug candidates for the identified target genes. The chemical structures of the drugs were obtained from PubChem in the structure data file format. The structures were transformed into PDB format using OpenBabel software (version 3.1.1) (http://openbabel.org)35. Protein crystal structures were sourced from the AlphaFold Protein Structure Database, which provides highly accurate protein structures comparable to those determined by advanced techniques, such as cryo-electron microscopy36. The protein receptors and small-molecule ligands were loaded into AutoDock Vina (version 4.2) for molecular docking37 and visualized using PyMOL software(https://pymol.org/2/). The ligand-receptor pairs with binding energies lower than − 1.2 kcal/mol were considered to have strong binding affinities, with lower values indicating better docking capability.

 

 

3. 

Zou M, Shao Z. Proteome-Wide Mendelian Randomization and Colocalization Analysis Identify Therapeutic Targets for Knee and Hip Osteoarthritis. Biomolecules. 2024;14(3):355. Published 2024 Mar 15. doi:10.3390/biom14030355

 

 

 

4.

Cao Z, Li Q, Li Y, Wu J. Identification of plasma protein markers of allergic disease risk: a mendelian randomization approach to proteomic analysis. BMC Genomics. 2024;25(1):503. Published 2024 May 22. doi:10.1186/s12864-024-10412-0