Skip to main content

Table 1 The 24 protein targets included in this study

From: Enhancing fragment-based protein structure prediction by customising fragment cardinality according to local secondary structure

#PDB IDLength% of helices% of strands% of coilsRelative size of the new 9-mer fileRelative size of the new 3-mer file
12CI2_I6516216385.8%51.7%
21CTF6851242585.2%49.7%
31DI26946332189.5%47.8%
41SCJ_B7123393886.5%49.5%
51HZ57230383279.2%37.0%
61CC87228353788.2%55.3%
73NZL735904179.5%41.4%
81DTJ7438263688.4%51.8%
91IG5756153483.5%47.6%
101OGW7625344181.4%47.9%
111DCJ8128244877.8%40.5%
121TIG8832323650.0%29.7%
131A198943174092.9%61.5%
141BM89937273693.9%58.4%
154UBP10054172977.2%46.5%
161IIB10355192694.9%73.4%
171M6T1067702384.8%50.1%
181ACF12534323477.8%45.1%
193CHY12845173873.9%45.3%
202KDL*566203887.7%50.2%
212LR8*705704374.0%44.0%
224HLB*9528244878.2%49.6%
232K4V*12528324072.1%48.4%
242KY4*1495914087.9%54.1%
Average88.742.320.737.082.1%49.0%
  1. * Not only did those targets prove particularly difficult to predict during CASP, but the Global Distance Test (GDT) of the best decoys (out of 20,000) in standard predictions is significantly lower (~ 50) than the GDT of the other targets of similar length and confirmed by Table ST1, the 5 added CASP targets create a more challenging set: the average GDT of their best decoy out of 20,000 (regardless of the energy score) is below 50 (i.e. 47), which is the threshold usually used to qualify an alignment as ‘good’ [34]. Moreover, as the table shows their associated first, best and average of the best 5 models are particularly poor for all targets except 2LR8