Table 4.

Comparison of probabilities that causal variants have rank 10 or less, from flashfm and fastPAINTOR, varying sample size.

	Trait 1 (A+D)				Trait 3 (I)
	Pr(1 or more cvs rank <= 10)		Pr(Both cvs rank <= 10)		Pr(rank cv <= 10)
N	flashfm	fastPAINTOR	flashfm	fastPAINTOR	flashfm	fastPAINTOR
1000	0.787	0.393	0.130	0.057	0.943	0.557
2000	0.890	0.733	0.563	0.243	0.983	0.750
3000	0.910	0.873	0.750	0.413	0.997	0.850
4000	0.937	0.917	0.837	0.560	0.997	0.903
5000	0.957	0.940	0.897	0.670	1	0.943

	Trait 2 (A+C+E)
	Pr(1 or more cvs rank <= 100)		Pr(2 or more cvs rank <= 10)		Pr(All 3 cvs rank < = 10)
N	flashfm	fastPAINTOR	flashfm	fastPAINTOR	flashfm	fastPAINTOR
1000	0.953	0.847	0.627	0.440	0.053	0.067
2000	1.000	0.963	0.910	0.747	0.497	0.253
3000	1.000	0.993	0.977	0.827	0.743	0.400
4000	1.000	0.993	0.987	0.923	0.863	0.527
5000	1.000	0.997	0.993	0.930	0.943	0.543

Flashfm tends to have higher probabilities than those from fastPAINTOR, especially for detecting all (multiple) causal variants of a trait. Three traits were simulated to have causal variants A+D, A+C+E, and I and trait correlation is 0.4. Sample size ranges from N = 1000 to 5000. The region has 345 SNPS and was simulated to mimic the LD structure of the IL2RA region, 10p-6030000-6220000 (GRCh37/hg19). Results are based on 300 replications.