Table of Contents
Some gory details of protein secondary structure prediction
PPT Slide
Goal of secondary structure prediction
Secondary structure predictions of 1. and 2. generation
Helix formation is local
b-sheet formation is NOT local
Problems of secondary structure predictions (before 1994)
Simple neural network
Training a neural network 1
Training a neural network 2
Training a neural network 3
Training a neural network 4
Neural networks classify points
Simple neural network with hidden layer
Neural Network for secondary structure
Secondary structure predictions of 1. and 2. generation
PPT Slide
PPT Slide
PPT Slide
Balanced training
PPT Slide
PHDsec: structure-to-structure network
Better prediction of segment lengths
Evolution has it!
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
Spectrin homology domain (SH3)
Prediction accuracy varies!
Why so bad?
Stronger predictions more accurate!
Correct prediction of correctly predicted residues
BAD errors are frequent!
False prediction for engineered proteins!
PHDsec: the un-g(l)ory details
Details PHDsec: Multiple alignment
PHDsec: the un-g(l)ory details
Details PHDsec: Multiple alignment
Secondary structure prediction
Secondary structure prediction 2000
Prediction of protein secondary structure
CAFASP statistics
CAFASP sec unique
CAFASP sec homologous
CAFASP concept
What is significant
Rank only if significant
EVA: automatic continuous EVAluation of structure prediction
EVA: automatic continuous EVAluation of structure prediction
EVA: secondary structure
EVA: secondary structure
Accuracy varies for proteins!
Averaging over many methods not always a good idea!
Some proteins predicted better
Reliability correlates with accuracy!
Conclusion
Availability of methods
|