Analysis of CASP8 predictions

We evaluated 'Server only' predictions and 'Human/Server' predictions using three scores: TS, TR and CS. The results are presented as sortable tables (need JavaScript enabled). Superpositions of models with target structures can be visualized in PyMOL by clicking on a score (need PyMOL installed). Targets are evaluated as whole chains and domains and are grouped in 5 categories by prediction difficulty. Separate evaluation tables are provied for each category.

Server only evaluation

Human/Server evaluation

Evaluation web pages are non-intuitive, but loaded with action. READ THIS:

• the design of this page is inspired by (=stolen from) the wonderful ROBETTA evaluation pages;
• mouse-click on TS, TR or CS field in the table header sorts the column in descending order;
• second mouse-click on the same TS, TR or CS field sorts the column in ascending order;
• mouse-click on the GROUP field sorts the groups alphabetically;
• mouse-over a score in the table shows the target number, server name and the model, this is handy for large tables and for the "Best", as you see the model number right away;
• mouse-click on a target number downloads a PyMOL script to display the rainbow-colored structure;
• mouse-double-click on a target number downloads the PDB file processed to match the target sequence. These processed PDBs are also available here;
• Ctrl-mouse-click on a target number downloads unmodified PDB;
• mouse-click on a TS, TR, or CS score downloads a PyMOL script to display a superposition of the target (green) and the model (cyan, yellow, orange and red for Cα within 1, 2, 4, and 8Å for TS and TR; and with residue fractional contribution more than 0.89, 0.74, 0.57 and 0.41 for CS, respectively, dark gray for regions not in the target structure, blue for residues contributing to "repulsion" of TR score);
• mouse-double-click on a TS, TR, or CS score downloads a PyMOL script to display the model colored as described above;
• mouse-click on a group name or number highlights the row, repeated click removes the highlight. This is handy to follow the group while re-sorting by different scores or targets.

WARNING: evaluation tables are generated automatically and do not represent the final word in CASP8 assessment. For other automatic evaluations see Baker, Cheng, McGuffin, Zhang.

questions, comments, suggestions: shuoyong.shi@utsouthwestern.edu

Target Categories in CASP8		First	Score from First model only
CM_easy:	Top 10 First-GTD_TS avg ≥ 81	Best	Score from Best scoring model (up to 5)
CM_medium:	Top 10 First-GDT_TS avg ≥ 67	Z-score	Z-score computed on the target scores
CM_hard:	Top 10 First-GDT_TS avg ≥ 52	TS	GDT_TS score by LGA 1)
FR:	Top 10 First-GDT_TS avg ≥ 30	TR	GDT_TS score by LGA 1) minus penalty
FM:	Top 10 First-GDT_TS avg < 30	CS	Contact score based on intramolecular distances
All Targets:	Everything combined	SUM	The sum of scores for all models

To display structures with a mouse-click

                                       

                                                               

Download: target seq | target str | domain def | evaluation scores