Table 1.

Bidirectional MR results for LST and MVPA with BMI or body fat percentage using significant loci only

Exposure	Outcome	Beta	s.e.	P value	Exposure	Outcome	Beta	s.e.	P value
LST	Body fat %	0.16	0.07	0.016	LST	BMI	0.40	0.04	8.4 × 10−14
Body fat %	LST	0.12	0.03	0.005	BMI	LST	0.16	0.01	1.4 × 10−74
MVPA	Body fat %	−0.21	0.17	0.22	MVPA	BMI	−0.25	0.04	0.002
Body fat %	MVPA	−0.001	0.036	0.97	BMI	MVPA	−0.10	0.01	5.8 × 10−12

We use MR-PRESSO with outliers removed for all pairs of traits except for the causal effect estimation between body fat percentage (body fat %) and MVPA because no outliers were detected by MR-PRESSO. For body fat percentage → MVPA, we reported the causal estimates using an inverse variance-weighted test; for MVPA → body fat percentage, we reported the weighted median method because these two methods were selected by the machine learning framework (Methods) to be the most appropriate approaches for each analysis, respectively. P < 0.0125 indicates significant effects.