Table 1.

Percentage of variation in depSympt^a attributable to the genotype‐covariate (G–C) interaction and the residual‐covariate (R–C) interaction with 95% confidence intervals (CIs), for covariate traits with significant interaction effects, showing the p value for the comparison of the full model to null model, with significance set at α = 0.05/17 $\approx$ 0.003

		Proportion of variability in depSympta attributable to
		G–C interaction (%)		R–C interaction (%)
Covariate	p value	Estimate	95% CI	Estimate	95% CI
Neuroticism	5.06E−139	−0.15	[−0.76, 0.46]	2.58	[  1.86, 3.30]
Childhood trauma	2.59E−058	0.59	[−0.14, 1.32]	2.98	[  2.18, 3.77]
Sleep	1.97E−041	1.22	[  0.54, 1.89]	2.52	[  1.78, 3.27]
BMI	6.36E−018	−0.23	[−0.86, 0.41]	1.39	[  0.68, 2.09]
Waist circumference	6.15E−016	−0.15	[−0.78, 0.48]	1.48	[  0.78, 2.19]
Smoking	2.49E−010	0.47	[−0.52, 1.46]	1.57	[  0.51, 2.63]
Waist‐to‐hip ratio	4.49E−009	−0.33	[−0.95, 0.29]	1.03	[  0.34, 1.73]
MET total	1.92E−007	0.23	[−0.42, 0.87]	0.53	[−0.17, 1.24]
MET walk	1.13E−005	0.10	[−0.55, 0.74]	1.18	[  0.45, 1.92]
MET mod	5.73E−004	−0.26	[−0.87, 0.35]	−0.08	[−0.78, 0.61]
TDI	1.96E−003	−0.19	[−0.81, 0.42]	1.67	[  0.97, 2.38]

Note: Polygenic and residual variance components for depSympt are functions of a covariate trait under the full MRNM. The percentage of variability in depSympt attributable to the G–C (R–C) interaction variance component relates to ${\sigma }_{{\alpha }_1}^2$ (${\sigma }_{{\tau }_1}^2$) in Equation (2). For a G–C interaction, a 95% CI that excludes zero shows that the covariate trait modulates polygenic effects on depSympt. For an R–C interaction, a 95% CI that excludes zero indicates that the covariate trait has some unmodelled relationship with depSympt.

^a Adjusted for age, sex, genetic batch and PCs 1 to 15 (SM section 1.4.4 provides details for producing interaction variance component estimates and standard errors on this scale).