. 2020 Apr 7;36(12):3758–3765. doi: 10.1093/bioinformatics/btaa234

Table 1.

Comparison of FASPR with four popular side-chain packing programs on the native structures from DB379, where bold fonts mark the best performer in each category

Method^a	$χ_{1 - 4}$ recovery rate (%)/RMSD (Å)			#Clash^b	Relative time^c
Method^a	All	Core	Surface	#Clash^b	Relative time^c
FASPR	69.1/1.457	80.3/0.983	56.8/1.906	149	1.00
CISRR	68.6/1.526	81.4/0.958	54.8/2.022	60	44.09
RASP	67.8/1.551	78.9/1.067	55.5/1.998	785	1.27
SCATD	61.7/1.856	74.7/1.279	48.4/2.318	1388	4.93
SCWRL4	68.8/1.524	80.7/0.966	55.5/1.991	557	26.77
SCWRL4v	67.0/1.620	78.9/1.061	54.1/2.072	232	5.11

SCWRL4 by default samples subrotamers and thus searches in a larger rotamer space; SCWRL4v runs the SCWRL4 program by disabling the subrotamers sampling with option “-v”. The other PSCP programs also use the Dunbrack rotamer library without sampling subrotamers.

#Clash, the total number of clashes for all the 379 packed structure models.

Reports how much slower on average the other methods are than FASPR. These values were calculated by averaging the ratio of the computational time at each column with respect to the FASPR column in Supplementary Table S10. All the programs were run on the XSEDE Comet server (Towns et al., 2014) using a single CPU [Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50 GHz].