A single-model quality assessment method for poor quality protein structure

BMC Bioinformatics

Table 2 Performance of various QA methods measured by GDT-TS and TM-scores (CASP13 FM domains, poor quality dataset)

QA methods	Corr. TM-score on stage2	Corr. GDT-TS on stage 2	Top 1 TM-score on stage 2 / Z-score sum	Top 1 GDT-TS on stage 2 / Z-score sum	Best model TM-score on stage 2 / Z-score sum	Best model GDT-TS on stage 2 / Z-score sum
Ours	0.79	0.80	44.91/45.43	39.58/49.17	51.33/66.94	44.16/66.47
DOPE	0.48	0.48	40.05/29.04	34.67/31.73	44.27/42.28	38.32/44.79
GOAP	0.40	0.42	33.70/11.66	28.94/11.87	42.77/39.61	36.89/40.07
ProQ4	0.47	0.43	32.70/13.74	27.87/12.80	43.93/45.57	37.70/45.93
ProQ3D	0.61	0.62	42.26/41.89	36.52/45.08	48.94/61.72	42.39/64.74
DeepQA	0.55	0.55	34.23/13.92	29.33/16.21	47.05/55.47	40.22/56.76

Z-score is calculated from the GDT-TS of the selected model and the GDT-TS of all models of the target. Sum the Z scores of each target to get the Z-score sum

ISSN: 1471-2105