Table 3. Different weighting schemes used in the protein folding simulations.

In the columns to the left, the number of threads, out of a total of 32, sampling structures below 2 and 4 Å C_α-RMSD respectively to the reference structure is shown. The sampled structures from each thread were divided into clusters and representative structures for each cluster were selected as the structure median in PROFASI+CamShift energy, from the 100 structures closest to the cluster centroid. The C_α-RMSD in Å of the lowest-energy cluster representative is shown below in the columns to the right.

	Threads (out of 32) sampling below 2Å (left) and 4Å (right)						Lowest-energy RMSD (Å)
	ENHD		Protein G		SMN		ENHD	Protein G	SMN
Gaussian/fixed	32	32	0	7	29	30	3.67	3.11	3.11
Gaussian/sampled	32	32	4	15	13	20	2.15	3.03	5.88
Gaussian/marginalized	32	32	1	16	7	14	4.24	2.72	6.06
Cauchy/fixed	32	32	9	25	15	21	1.94	1.15	2.58
Cauchy/sampled	32	32	13	24	11	16	1.87	2.82	5.51
Square well/α = 1a	19	22	2	12	14	18	2.29	3.14	3.71
Square well/α = 5a	32	32	0	1	1	5	3.82	5.83	1.91
CS-Torus b	4	27	8	25	0	0	19.2	3.01	8.33

Notes.

^aWeights, α, of 1 and 5 were used by Robustelli et al.

^bLowest-energy cluster representatives for the CS-Torus simulations were selected from PROFASI energy alone.