Table 3. The probabilistic multi-template modeling approach is less negatively affected by bad templates.

Mean GDT-ha scores of 1000 models built with templates sets containing 0, 1 and 2 bad templates (TMscore < 0.3) along with two good templates (TMscore > 0.5).

Good templates	Bad templates	Modeller restraints	New restraints
2	0	0.474 (-)	0.480 (-)
2	1	0.466 (−1.7%)	0.475 (−1.0%)
2	2	0.458 (−3.4%)	0.471 (−1.9%)