Variance components models for gene--environment interaction in twin and sib-pair linkage analysis analysis

Shaun Purcell

Introduction | Mx scripts | gxe-visual.exe | Mx
These Mx scripts are intended to facilitate gene-environment interaction in twin and sib-pair analysis, specifically when using continuous moderator variables, as discussed in these two manuscripts. Also, a special program for plotting variance components as a function of a moderator variable can be downloaded, gxe-visual. Please reference the articles if you use the scripts in a publication.

Mx Scripts
If you have technical problems running these scripts after reading the associated manuscripts and studying the scripts themselves, please let me know. The more complex models will often require good starting values and multiple attempts at convergence. All scripts were developed using the 1.47c Linux version of Mx.

Simple GxE
File :
Notes : Datafile must be in the format
ID ZYG T1 T2 M1 M2
Fix parameters T, U, V and B to test for moderated effects on A, C, E and M respectively.

Nonlinear GxE
File :
Notes : As above, but with extra parameters X, Y and Z representing quadratic interaction terms. Use multiple start values to ensure a global minimum has been found.

Two moderator GxE
File :
Notes : Two (linear) moderating variables. Data format is
where MA is the first moderator and MB is the second moderator (1,2 refers to twin 1, twin 2)

GxE in presence of r_GE
File : and
Notes : Data format for must be
ZYG T1 T2 M1 M2 M1 M2
i.e. where the same moderator variable is actually entered twice. The parameters T and U are the interaction parameters for the common and unique paths.

The first script represents the model as described in the Twin Research article and only allows for moderation of the additive genetic component. The second script applies the same model to the C and E components and should be used in preference to the first.

Scalar and qualitative GxE
File :
Notes : Data format as for The A parameter is the first, unmoderated polygenic component. The T and O parameters are the second moderated polygenic component.

File :
Notes : Data format must be
ID ZYG T1 T2 M1 M2 Z0 Z1 Z2 
where Z0, Z1 and Z2 are the estimated IBD probabilities at the test locus. The parameter Q estimates QTL variance; C and E estimate shared and nonshared residual components. P estimates moderated QTL effects; U and V represent moderated shared and nonshared residual effects.

QTL incorporating ExE
File :
Notes : Data format must be
T1 T2 Z0 Z1 Z2 R D1 D2 
where Z0, Z1 and Z2 are IBD probabilities, R is the imputed pair=specific correlation and D1 and D2 are simply `dummy' columns full of 0's (zeros).

File :
Notes : This file is an excerpt from a script to create the imputed pair-specific correlation dataset based on prior estimates of moderated parameters. It is only intended as an illustrative example -- it would need to be changed somewhat for different datasets.

GxE Visualisation tool (gxe-visual)
A MS Windows 95/98/2000/NT program to help visualise continuously moderated variance components, with documentation

Mx is the model-fitting package used to implement the DF model. It can be freely downloaded for various platforms.

Purcell S (2002). Variance components models for gene-environment interaction in twin analysis. Twin Research, 5(6), 554-571.

Purcell S & Sham PC (2002). Variance components models for gene-environment interaction in quantitative trait locus linkage analysis. Twin Research, 5(6), 572 - 576.
Page created by Shaun Purcell.