Table 2.

Linear regression analyses testing lipid polygenic risk score (PRS) as a predictor of its corresponding lipid level, in a univariable model and in a multivariable model accounting for phenotypic, batch, and ancestry variables

	TG (n = 149)a			HDLC (n = 150)			LDLC (n = 148)b			TC (n = 151)b
	beta	std error	p	beta	std error	p	beta	std error	p	beta	std error	p
Univariable model
PRS (for the lipid level assessed)	0.2403	0.0803	0.0032	0.4579	0.0737	5.00E-09	0.2403	0.0805	0.0034	0.2262	0.0809	0.0058
Multivariable model
PRS (for the lipid level assessed)	0.3133	0.0804	1.52E-04	0.3818	0.0667	6.28E-08	0.2524	0.0839	0.0031	0.2494	0.0851	0.0040
Female sex	-0.2633	0.0771	8.36E-04	0.3980	0.0669	2.17E-08	0.0717	0.0862	0.4067	0.0818	0.0853	0.3392
Age	0.0019	0.0849	0.9820	0.0602	0.0721	0.4050	0.0777	0.0940	0.4099	0.1237	0.0939	0.1899
BMI (kg/m2)	0.2944	0.0803	3.51E-04	-0.3009	0.0673	1.61E-05	0.1013	0.0881	0.2522	0.1040	0.0881	0.2398
T2D	-0.0642	0.0819	0.4343	0.0590	0.0697	0.3990	-0.0328	0.0928	0.7247	-0.0001	0.0908	0.9989
Psychotic illnessc	0.0472	0.0792	0.5527	-0.0025	0.0669	0.9710	0.0347	0.0885	0.6957	0.0068	0.0870	0.9380
Originating ascertainment cohort	-0.0676	0.1559	0.6654	0.1313	0.1329	0.3250	0.0608	0.1727	0.7255	-0.0471	0.1728	0.7857
Sequencing platform	0.1972	0.1230	0.1112	-0.1072	0.1047	0.3080	-0.0629	0.1388	0.6512	0.0402	0.1369	0.7697
Ancestry PC1	-0.0827	0.0787	0.2953	0.0078	0.0663	0.9060	-0.0659	0.0870	0.4503	-0.0476	0.0870	0.5852
Ancestry PC2	-0.1035	0.0989	0.2970	0.1073	0.0823	0.1950	0.0811	0.1073	0.4509	0.0823	0.1072	0.4439
Ancestry PC3	0.1311	0.1407	0.3532	-0.0049	0.1195	0.9680	-0.0847	0.1555	0.5868	-0.1042	0.1557	0.5043
Ancestry PC4	0.0366	0.0853	0.6687	-0.0755	0.0730	0.3030	-0.0808	0.0942	0.3926	-0.1176	0.0944	0.2151
	R2		p	R2		p	R2		p	R2		p
Model	0.2471		7.26E-05	0.4565		2.46E-13	0.0914		0.3400	0.0909		0.3258
	ΔR2			ΔR2			ΔR2			ΔR2
PRS variable	0.0843			0.1302			0.0647			0.0630

Beta coefficients are standardized with positive values indicating a positive association between higher lipid levels and female sex, older age, higher BMI, having type 2 diabetes, having a psychotic illness, belonging to the TCAG sequencing project cohort, sequenced using a HiSeq X, and higher value for PC (and thus a negative beta value indicates a negative association for that variable, i.e., lower lipid level)

ΔR², here an estimate of the variance in each lipid level explained by the PRS, is calculated by subtracting the R² of the multivariable model without the PRS variable (i.e., using all remaining phenotypic, batch, and ancestry variables) from the R² of the full model

TG Triglyceride, LDLC Low density lipoprotein cholesterol, HDLC High density lipoprotein cholesterol, TC Total cholesterol, BMI Body mass index, T2D Type 2 diabetes, PC Principal component for ancestry

Bold font indicates statistical significance

^aExcluded one individual on fibrate treatment. Triglyceride levels were natural log transformed to approximate a normal distribution

^bFor individuals on statin medications, LDLC and TC levels were divided by 0.7 and 0.8, respectively, as in previous studies [8, 9, 32]

^cDefined as schizophrenia or schizoaffective disorder