Figures in the main text
- Figure 1: illustion
  * rationale of variance loci
  * existing tests (i.e., DLM) fail on binary
  * VLA fits a 2nd-order polynomial curve and test positive gradient
  * existing tests (i.e., DLM) is less efficient under pure GxE

- Figure 2: Selected Simulation Results
  * display false positive rate (when GxE is null)
  * display net positive rate (when GxE is not null)
  * shows Gaussian, balance binary, and imbalanced binary. (do not show Poisson)

- Figure 3: Manhatton and QQ Plots for whole genome putative GxE tests
  * 3a: on BMI
  * 3b: on T2D

- Figure 4: enrichment analysis
  * 4a: internal (Putative GxE based on British Whites --> empirical GxE based on non-British Whites)
  * 4b: external (Putative GxE based on UKBB Whites --> empirical GxE based on PEGS participants)
  * 4c: functional (Putative GxE based on UKBB Whites --> variant functional labels by ANNOVAR)

- Figure 5: GxE discovery around ABCB7