Talk:Mendelian Randomization: Difference between revisions

From Bridges Lab Protocols
Jump to navigationJump to search
Davebrid1 (talk | contribs)
Bureaucrats, Administrators
21 edits
Added notes on what to add to MR SOP
 
added notes about handling outliers
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:
== MR SOP — outstanding items (selection/collider bias work, 2026-07-18) ==


## A. MR SOP — outstanding items
Context: §4 "Selection and Collider Bias" was added to the [https://bridgeslab.sph.umich.edu/protocols/index.php/Mendelian_Randomization MR SOP] after the smoking→AD project found its headline result was most plausibly a survival-collider artefact.


### Sections still missing (all cheap, all bit us on smoking→AD)
=== Sections still missing ===
1. **Steiger directionality / reverse causation** — absent entirely. One line
All three are cheap, and all three bit us on smoking→AD.
  (`TwoSampleMR::directionality_test()`). Important for the disease→AD portfolio, where
  reverse causation is plausible.
2. **Sample overlap** — absent. Two-sample MR assumes independent samples; UK Biobank is in
  nearly every GWAS we use (it was in *both* our exposure and outcome). Tool: MRlap
  (needs full genome-wide sumstats + LD scores).
3. **Dataset provenance verification** — absent. We caught **two mislabelled OpenGWAS
  accessions**: `ieu-a-1001` labelled "parental lifespan" is actually Okbay EA;
  `ieu-b-5067` labelled "Wightman" is Woolf 2022 with 954 cases. Rule: verify N,
  case/control split and cohort composition via `gwasinfo()` before use; never trust the label.


### MR-PRESSO / MR-Egger imbalance (raised, not yet drafted)
# '''Steiger directionality / reverse causation''' — absent entirely. One line (<code>TwoSampleMR::directionality_test()</code>). Important for the disease→AD portfolio, where reverse causation is plausible.
Observation: we report MR-PRESSO outlier-corrected far more than MR-Egger. This is
# '''Sample overlap''' — absent. Two-sample MR assumes independent samples; UK Biobank is in nearly every GWAS we use (it was in ''both'' our exposure and outcome). Tool: MRlap (needs full genome-wide sumstats + LD scores).
**structural**, not accidental — Egger intercept tests are underpowered so Scenario 4 rarely
# '''Dataset provenance verification''' — absent. We caught '''two mislabelled OpenGWAS accessions''': <code>ieu-a-1001</code> labelled "parental lifespan" is actually Okbay EA; <code>ieu-b-5067</code> labelled "Wightman" is Woolf 2022 with 954 cases. Rule: verify N, case/control split and cohort composition via <code>gwasinfo()</code> before use; never trust the label.
fires, while Cochran's Q is almost always significant so Scenario 2 fires constantly.


Why it matters — our own Layer 0: IVW b=−0.065 **p=0.296** → MR-PRESSO b=−0.112 **p=0.0015**.
=== MR-PRESSO / MR-Egger imbalance (raised, not yet drafted) ===
Outlier removal *manufactured* the significant result later shown to be collider-driven. It
Observation: we report MR-PRESSO outlier-corrected far more often than MR-Egger. This is '''structural''', not accidental — Egger intercept tests are underpowered so Scenario 4 rarely fires, while Cochran's Q is almost always significant so Scenario 2 fires constantly.
removed only 2/22 SNPs (9%), so the existing ">30% removed" tripwire never fired.


Proposed amendments (Scenario 2 currently has **no** conclusion-change tripwire):
Why it matters — our own Layer 0: IVW b=−0.065 '''p=0.296''' → MR-PRESSO b=−0.112 '''p=0.0015'''. Outlier removal ''manufactured'' the significant result later shown to be collider-driven. It removed only 2/22 SNPs (9%), so the existing ">30% removed" tripwire never fired.
- Add a **conclusion-change tripwire** to Scenario 2: if outlier removal moves the estimate
 
  outside the original IVW 95% CI or flips significance → report corrected + uncorrected,
Proposed amendments (Scenario 2 currently has '''no''' conclusion-change tripwire):
  corrected as sensitivity only. (Key on *effect on conclusion*, not SNP count.)
* Add a '''conclusion-change tripwire''' to Scenario 2: if outlier removal moves the estimate outside the original IVW 95% CI or flips significance → report corrected + uncorrected, with the corrected value as sensitivity only. Key on ''effect on the conclusion'', not SNP count.
- Stop treating **Egger intercept p>0.05** as evidence of no directional pleiotropy — it's an
* Stop treating '''Egger intercept p>0.05''' as evidence of no directional pleiotropy — it is an underpowered null, the same error pattern we rejected in §4.2. Report the intercept estimate with CI.
  underpowered null, the same error pattern we rejected in §4.2. Report intercept + CI.
* Elevate '''Weighted Median''' to co-primary (better power ''and'' robustness than MR-Egger; currently only "support" in Scenario 2).
- Elevate **weighted median** to co-primary (better power *and* robustness than Egger; currently
* Report the '''full estimator panel with a concordance flag''' — our estimates spanned −0.065 to −0.208 with p from 0.0015 to 0.647; single-winner adjudication hides that spread.
  only "support" in Scenario 2).
 
- Report the **full estimator panel with a concordance flag** — our estimates spanned
 
  −0.065 to −0.208 with p from 0.0015 to 0.647; single-winner adjudication hides that.
For handling sample overlap, some options, especially when using UKBB
* MR-APSS
* MRBEE
* CAUSE
 
For modeling correlated horizontal pleiotropy
* MR-APSS does this by background correction
* CAUSE does this by separating and modeling correlated horizontal pleiotropy

Latest revision as of 12:24, 23 October 2026

MR SOP — outstanding items (selection/collider bias work, 2026-07-18)

Context: §4 "Selection and Collider Bias" was added to the MR SOP after the smoking→AD project found its headline result was most plausibly a survival-collider artefact.

Sections still missing

All three are cheap, and all three bit us on smoking→AD.

  1. Steiger directionality / reverse causation — absent entirely. One line (TwoSampleMR::directionality_test()). Important for the disease→AD portfolio, where reverse causation is plausible.
  2. Sample overlap — absent. Two-sample MR assumes independent samples; UK Biobank is in nearly every GWAS we use (it was in both our exposure and outcome). Tool: MRlap (needs full genome-wide sumstats + LD scores).
  3. Dataset provenance verification — absent. We caught two mislabelled OpenGWAS accessions: ieu-a-1001 labelled "parental lifespan" is actually Okbay EA; ieu-b-5067 labelled "Wightman" is Woolf 2022 with 954 cases. Rule: verify N, case/control split and cohort composition via gwasinfo() before use; never trust the label.

MR-PRESSO / MR-Egger imbalance (raised, not yet drafted)

Observation: we report MR-PRESSO outlier-corrected far more often than MR-Egger. This is structural, not accidental — Egger intercept tests are underpowered so Scenario 4 rarely fires, while Cochran's Q is almost always significant so Scenario 2 fires constantly.

Why it matters — our own Layer 0: IVW b=−0.065 p=0.296 → MR-PRESSO b=−0.112 p=0.0015. Outlier removal manufactured the significant result later shown to be collider-driven. It removed only 2/22 SNPs (9%), so the existing ">30% removed" tripwire never fired.

Proposed amendments (Scenario 2 currently has no conclusion-change tripwire):

  • Add a conclusion-change tripwire to Scenario 2: if outlier removal moves the estimate outside the original IVW 95% CI or flips significance → report corrected + uncorrected, with the corrected value as sensitivity only. Key on effect on the conclusion, not SNP count.
  • Stop treating Egger intercept p>0.05 as evidence of no directional pleiotropy — it is an underpowered null, the same error pattern we rejected in §4.2. Report the intercept estimate with CI.
  • Elevate Weighted Median to co-primary (better power and robustness than MR-Egger; currently only "support" in Scenario 2).
  • Report the full estimator panel with a concordance flag — our estimates spanned −0.065 to −0.208 with p from 0.0015 to 0.647; single-winner adjudication hides that spread.


For handling sample overlap, some options, especially when using UKBB

  • MR-APSS
  • MRBEE
  • CAUSE

For modeling correlated horizontal pleiotropy

  • MR-APSS does this by background correction
  • CAUSE does this by separating and modeling correlated horizontal pleiotropy
Retrieved from "https://bridgeslab.sph.umich.edu/protocols/index.php?title=Talk:Mendelian_Randomization&oldid=2787"