Analysis1
1. Read the important SNPs
2. Superwindow removal
3. Variance component estimation 
4. OLS on the pairwise-interaction and get the coefficient and p-values


Analysis1.5
1. Read the important SNPs
2. Superwindow removal
3. Variance component estimation 
4. OLS on the pairwise-interaction and get the coefficient and p-values
5. Bayesian Posterior on the pairwise-interaction and get coefficietn and p-values


Analysis2:
1. 10KB upstream and downstream for each gene --> same analysis as analysis 1 


Analysis3:
1. Comparison between main variance component and the quad-effect


Analysis 4:
Run the same experiments with 40 PC removal

Analysis 5: 
Run the same experiments on the imputed genotype data

Analysis 6:
Imputed data has features too correlated. Maybe should try including the linear effect



### May 20th, 2024 ###
Analysis 5.5
Run the same experiments on the imputed genotype data -- with high VIF feature removal


### Jun 3rd, 2024 ###
sub1.5.v2.sh: fixed the bug on matrix index shuffling issue
quadkast_sig_genes.v2.csv: Prateek additionally detected signals for billirubin_total 