Skip to content

Latest commit

 

History

History
311 lines (198 loc) · 14 KB

Bidirectional Mendelian Randomization Lone and F1.md

File metadata and controls

311 lines (198 loc) · 14 KB
Raw

Bidirectional MR

loneliness to F1/NMD

F1 refers to the Neurodevelopmental and Mood Disorders latent factor (NMD)

  • clumping parameters: r2 < 0.001, kb = 10,000
  • p-value threshold: p < 5x10-08
  • 11 instruments

Results from two sample MR:

outcome exposure method nsnp b se pval
F1 loneliness MR Egger 11 0.5443916 0.2789841 8.28x10-02
F1 loneliness Inverse variance weighted 11 0.4956359 0.0547951 1.49x10-19
F1 loneliness Weighted mode 11 0.5238836 0.1343346 2.36x10-03
F1 loneliness Weighted median 11 0.4810642 0.0803782 7.22x10-09
  • All methods except for Egger are significant

Heterogeneity tests

id.exposure id.outcome outcome exposure method Q Q_df Q_pval
zekZiO bCloOr F1 loneliness MR Egger 8.840471 9 0.4521287
zekZiO bCloOr F1 loneliness Inverse variance weighted 8.872238 10 0.5442702
  • since Q is low and not statistically significant, there is no evidence of heterogeneity

Test for directional horizontal pleiotropy

id.exposure id.outcome outcome exposure egger_intercept se pval
zekZiO bCloOr F1 loneliness -0.0007018 0.0039373 0.8624872
  • test for pleiotropy suggests no horizontal pleiotropy

Test that the loneliness exposure is upstream of the F1 outcome

exposure outcome snp_r2.exposure snp_r2.outcome correct_causal_direction steiger_pval
loneliness F1 0.0009896 0.0002397 TRUE 4.87x10-13
  • Steiger directionality test suggests that we have the correct direction of causal effect from loneliness to F1

MR Egger Sensitivity Test

outcome exposure method nsnp b se pval
F1 loneliness MR Egger 77 1.4606595 0.3571151 1.07x10-04
F1 loneliness Inverse variance weighted 77 0.4687211 0.0449261 1.75x10-25
F1 loneliness Weighted mode 77 0.6736313 0.1070706 1.84x10-08
F1 loneliness Weighted median 77 0.5464386 0.0363076 3.44x10-51
  • still significant and consistent across MR methods
  • I2 of all instruments = .972, I2 of weaker instruments .972
  • not suggestive of weak instrument bias

Forest plot of single SNP MR

image

Scatter Plot

image

Steiger Filtering

  • all SNPs were found to be operating in the correct causal direction
  • so results of MR steiger are the same as normal MR

F1 QSNP filtered to Loneliness

  • clumping parameters: r2 < 0.001, kb = 10,000
  • p-value threshold: p < 5x10-08
  • 11 instruments

Results from Two Sample MR

outcome exposure method nsnp b se pval
loneliness F1_Qsnp MR Egger 10 0.4439023 0.3011940 1.79x10-01
loneliness F1_Qsnp Inverse variance weighted 10 0.3393446 0.0654521 2.16x10-07
loneliness F1_Qsnp Weighted mode 10 0.3857237 0.0880167 2.61x10-03
loneliness F1_Qsnp Weighted median 10 0.3826074 0.0639450 3.60x10-09
  • MR methods are significant except for MR Egger

Heterogeneity Test

outcome exposure method Q Q_df Q_pval
loneliness F1_Qsnp MR Egger 20.81972 8 0.0076421
loneliness F1_Qsnp Inverse variance weighted 21.15064 9 0.0119981
  • there is evidence of heterogeneity

Test for Directional Horizontal Pleiotropy

outcome exposure egger_intercept se pval
loneliness F1_Qsnp -0.0017578 0.0049295 0.7306186
  • no evidence of directional horizontal pleiotropy

MR Egger Sensitivity Test

id.exposure id.outcome outcome exposure method nsnp b se pval
AzksNX 7Tkurt loneliness F1_Qsnp MR Egger 91 0.3000644 0.1401681 0.0350286
AzksNX 7Tkurt loneliness F1_Qsnp Inverse variance weighted 91 0.2963937 0.0230410 0.0000000
AzksNX 7Tkurt loneliness F1_Qsnp Weighted mode 91 0.3596411 0.0527016 0.0000000
AzksNX 7Tkurt loneliness F1_Qsnp Weighted median 91 0.3189740 0.0213165 0.0000000
  • I2 all instruments = .971, weak instruments = .970

Test that F1/NMD QSNP is upstream of loneliness

exposure outcome snp_r2.exposure snp_r2.outcome correct_causal_direction steiger_pval
F1_Qsnp loneliness 0.0011072 0.0002447 TRUE 1.49x10-15

Forest Plot

image

Scatter Plot

image

MR Steiger

  • all SNPs were found to be operating in the correct causal direction
  • so results of MR Steiger are identical to that of the normal MR

Loneliness to F1/NMD with reduced p-value threshold

  • clumping parameters: r2 < 0.001, kb = 10,000
  • p-value threshold: p < 5x10-07
  • 24 instruments

Results from Two Sample MR

outcome exposure method nsnp b se pval
F1 loneliness MR Egger 24 0.2901209 0.3827217 4.56x10-01
F1 loneliness Inverse variance weighted 24 0.4898733 0.0741826 4.01x10-11
F1 loneliness Weighted mode 24 0.5341551 0.1095374 5.93x10-05
F1 loneliness Weighted median 24 0.4828668 0.0653770 1.24x10-13
  • MR methods significant and consistent apart from MR Egger

Hetereogeneity

outcome exposure method Q Q_df Q_pval
F1 loneliness MR Egger 74.76671 22 1.15x10-07
F1 loneliness Inverse variance weighted 75.72983 23 1.54x10-07
  • evidence of heterogeneity

Horizontal Directional Pleiotropy

outcome exposure egger_intercept se pval
F1 loneliness 0.0026415 0.004962 0.5998197
  • no evidence of directional horizontal pleiotropy

MR Egger Sensitivity Analysis

outcome exposure method nsnp b se pval
F1 loneliness MR Egger 139 0.7990098 0.2366442 0.0009559
F1 loneliness Inverse variance weighted 139 0.4228029 0.0337056 0.0000000
F1 loneliness Weighted mode 139 0.6591925 0.1130996 0.0000000
F1 loneliness Weighted median 139 0.4874169 0.0289665 0.0000000
  • I2 of all instruments = 0.970, I2 of weak instruments = 0.970, so there does not seem to be weak instrument bias

Forest Plot

image

Scatter Plot

image

Test that loneliness is upstream of F1/NMD

exposure outcome snp_r2.exposure snp_r2.outcome correct_causal_direction steiger_pval
loneliness F1 0.0017882 0.0005796 TRUE 0
  • 2 instruments were found to be operating in the incorrect direction

MR Steiger

outcome exposure method nsnp b se pval
F1 loneliness MR Egger 22 0.4625481 0.3004663 0.1393710
F1 loneliness Inverse variance weighted 22 0.4355445 0.0583633 8.48x10-14
F1 loneliness Weighted mode 22 0.5412482 0.1115567 8.51x10-05
F1 loneliness Weighted median 22 0.4741883 0.0645738 2.08x10-13
  • results are still consistent with the original MR analysis, except that the MR Egger Bootstrapped effect is no longer in the negative direction

MR RAPS

  • beta hat (causal effect) = 0.5128462
  • beta se = 0.06853184
  • beta p value = 7.238654e-14

NMD/F1 Qsnp to Loneliness with Reduced p-value threshold

  • clumping parameters: r2 < 0.001, kb = 10,000
  • p-value threshold: p < 5x10-07
  • 19 instruments

Two Sample MR Results

outcome exposure method nsnp b se pval
loneliness F1_Qsnp MR Egger 19 0.2059939 0.2780956 4.69x10-01
loneliness F1_Qsnp Inverse variance weighted 19 0.3466251 0.0548549 2.63x10-10
loneliness F1_Qsnp Weighted mode 19 0.3661856 0.0769367 1.58x10-04
loneliness F1_Qsnp Weighted median 19 0.3358195 0.0520162 1.53x10-10
  • significant and consistent results across MR methods except for MR Egger

Heterogeneity

outcome exposure method Q Q_df Q_pval
loneliness F1_Qsnp MR Egger 48.75725 17 6.57x10-05
loneliness F1_Qsnp Inverse variance weighted 49.52169 18 8.91x10-05
  • there is evidence of heterogeneity

Directional Horizontal Pleiotropy

outcome exposure egger_intercept se pval
loneliness F1_Qsnp 0.0021538 0.0041719 0.6123145
  • no evidence of directional horizontal pleiotropy

Egger Sensitivity Analysis

outcome exposure method nsnp b se pval
loneliness F1_Qsnp MR Egger 146 0.2111061 0.0998598 0.0362366
loneliness F1_Qsnp Inverse variance weighted 146 0.2835963 0.0185522 9.42x10-53
loneliness F1_Qsnp Weighted mode 146 0.3703243 0.0520006 4.59x10-11
loneliness F1_Qsnp Weighted median 146 0.3106336 0.0191519 3.67x10-59
  • I2 all instruments = .964, I2 weak instruments = .962, so no weak instrument bias

Testing that F1 QSNP is upstream of loneliness

exposure outcome snp_r2.exposure snp_r2.outcome correct_causal_direction steiger_pval
F1_Qsnp loneliness 0.0018235 0.000444 TRUE 0

Forest Plot

image

Scatter Plot

image

MR Steiger

  • 1 SNP found to be operating in the incorrect direction
outcome exposure method nsnp b se pval
loneliness F1_Qsnp MR Egger 18 0.4306750 0.2357781 0.0864743
loneliness F1_Qsnp Inverse variance weighted 18 0.3222953 0.0448503 6.67x10-13
loneliness F1_Qsnp Weighted mode 18 0.3650974 0.0829786 0.0003913
loneliness F1_Qsnp Weighted median 18 0.3330525 0.0503948 3.87x10-11
  • even though the effect estimates have increased slightly, the pattern of results is mostly the sam

MR Raps

  • MR Raps indicates a significant causal effect from F1 Qsnp to Loneliness

  • beta hat (causal effect) = .358

  • beta se = 0.0497

  • beta p value = 1.33e-15