Skip to main content

Table 4 Comparison of 14 model quality assessment methods and their consensus ranking

From: A large-scale conformation sampling and evaluation server for protein tertiary structure prediction and its assessment in CASP11

QA method Best of top 1 Avg loss Best in the pool Avg loss
Consensus ranking 34 0.04 17 0.07
SELECTpro 32 0.05 17 0.08
ModFOLDclust2 30 0.07 18 0.10
APOLLO 30 0.07 16 0.09
Pcons 29 0.07 16 0.10
ProQ2 27 0.05 15 0.07
QApro 18 0.07 8 0.09
ModelCheck2 16 0.16 10 0.18
MULTICOM-NOVEL_QA 11 0.11 4 0.14
DFIRE2 9 0.11 6 0.14
Dope 9 0.11 6 0.14
ModelEvaluator 9 0.13 6 0.16
OPUS_PSP 9 0.11 6 0.14
RF_CB_SRS_OD 9 0.11 6 0.14
RWplus 9 0.11 6 0.14
  1. “Best of top 1” means the number of times when top 1 models selected by an individual QA method were actually the best of the top 1 models identified by all the QA methods. “Best in the pool” means the number of times when top 1 models by an individual method were actually the best models in the MULTICOM model pool. “Avg loss” means the average loss of GDT-TS scores between the best models and top 1 models ranked by each QA method