Skip to main content

Table 2 Comparison of Evo-Diverse to other algorithms on lowest energy via 1-sided Fisher’s and Barnard’s tests on the benchmark dataset. Top panel evaluates the null hypothesis that Evo-Diverse does not achieve the lowest energy, considering each of the other four algorithms in turn. The bottom panel evaluates the null hypothesis that Evo-Diverse does not achieve a lower lowest energy value in comparison to a particular algorithm, considering each of the four other algorithms in turn. Comparison of Evo-Diverse to other algorithms on lowest lRMSD via 1-sided Fisher’s and Barnard’s tests on the benchmark dataset. Top panel evaluates the null hypothesis that Evo-Diverse does not achieve the lowest lRMSD, considering each of the other four algorithms in turn. The bottom panel evaluates the null hypothesis that Evo-Diverse does not achieve a lower lowest lRMSD value in comparison to a particular algorithm, considering each of the four other algorithms in turn

From: Balancing multiple objectives in conformation sampling to control decoy diversity in template-free protein structure prediction

Test

mEA

mEA-PR

mEA-PR+PC

Rosetta

(a)

 Best lowest energy

Fisher’s

0.04118

0.088

0.04118

0.004181

Barnard’s

0.02489

0.05368

0.02489

0.001879

 Better lowest energy

Fisher’s

0.01282

0.3762

0.1715

0.00018

Barnard’s

0.008299

0.3179

0.1341

0.00009139

(b)

 Best lowest lRMSD

Fisher’s

0.001671

0.006907

0.001671

0.5

Barnard’s

0.000702

0.003284

0.000702

0.4373

 Better lowest lRMSD

Fisher’s

0.001924

0.01282

0.001924

0.3762

Barnard’s

0.001118

0.008299

0.001118

0.3179

  1. p-values less than 0.05 are marked in bold