Negative Genetic correlation in Common Factor GWAS

[Yu - Yvonne Huang]

Dear All,

I'm running a CommonFactorGWAS for 4 related traits. The LDSC result shows some traits have negative genetic correlations with others.

Genetic Correlation between CRAE_UKB and AVR_UKB: 0.3179 (0.1664)     
Genetic Correlation between CRAE_UKB and SBP_EVAN: -0.2648 (0.0462)     
Genetic Correlation between CRAE_UKB and DBP_EVAN: -0.3031 (0.0468)     
Genetic Correlation between AVR_UKB and SBP_EVAN: -0.3144 (0.0603)     
Genetic Correlation between AVR_UKB and DBP_EVAN: -0.346 (0.0607)     
Genetic Correlation between SBP_EVAN and DBP_EVAN: 0.8099 (0.0392)

 I run the CFA model with one factor, and the model fits well:

CommonFactor_ML
$modelfit
      chisq df    p_chisq      AIC       CFI       SRMR
df 5.576171  2 0.06153892 21.57617 0.9995867 0.06217045

$results
       lhs op      rhs Unstandardized_Estimate Unstandardized_SE
1       F1 =~ CRAE_UKB              0.11715047       0.017421367
2       F1 =~  AVR_UKB              0.09745155       0.016105531
3       F1 =~ SBP_EVAN             -0.32194645       0.015622345
4       F1 =~ DBP_EVAN             -0.34598825       0.018050336
6 CRAE_UKB ~~ CRAE_UKB              0.11487112       0.018135033
7  AVR_UKB ~~  AVR_UKB              0.05827442       0.016975698
8 SBP_EVAN ~~ SBP_EVAN              0.03626659       0.009312294
9 DBP_EVAN ~~ DBP_EVAN              0.01589492       0.010970629

  Standardized_Est Standardized_SE      p_value
1        0.3266867      0.04858135 1.761628e-11
2        0.3743399      0.06186605 1.441171e-09
3       -0.8606960      0.04176498 2.322930e-94
4       -0.9395655      0.04901747 6.853501e-82
6        0.8932758      0.14102402 2.385594e-10
7        0.8598696      0.25048534 5.973511e-04
8        0.2592024      0.06655626 9.840754e-05
9        0.1172168      0.08090267 1.473763e-01

Then I ran:

 VESSELWIDTH_factor <- commonfactorGWAS(covstruc = LDSCoutput, SNPs = p_sumstats1, estimation = "ML", cores = 20, parallel = TRUE,smooth_check=TRUE)

In independent GWAS, each of the traits has several significant hits. However, the CommonFactorGWAS (F1 ~ SNP) did not give any significant hits....

Can this happen?  I assume extracting the common factor from different traits will be similar to conducting MTAG, which will boost the power of GWAS...But in my case, it didn't...

Best wishes,

Yvonne