Chen, P., Deane, C.M., Reinert G. 2008. Submitted
Reference : P. Chen (2005) A Bayesian approach to predicting protein-protein interactions. D.Phil transfer report, Deprtment of Statistics, Oxford University. [report]
Programs (*.pyc) are compiled using Python 2.4.
Run PYC files. Please follow the popup questions and input the corresponing filenames for a successful prediction.
Building upcast sets of triples (triangles, lines and triples) of characteristic categories
Use triple-wise protein interactions and protein annotation
Include triples (triples, triangles, lines) of characteristic categories
Integrate multiple characteristics
Query protein pairs
A list of protein pairs to be predicted.
As an eligible protein pair (predictable), the characteristics, i.e. structure and/or function, of the two flanking proteins are known and there exists at least one common interacting protein neighbours. For the common interacting proteins, the characteristics are also known.
Method
The triangle rate score
Example -- Predicting 5 selected protein pairs
Sample datasets
Query protein pairs (eligible protein pairs)
Convert classifications from txt file to Python shelve PYC file
Output -- structural classification (shelve), functional classification (shelve)
Construct upcast set of triples (triangles and lines) of characteristic categoriesPYC file
Output -- use both structure and function classifcations (triangles, lines)
Prediction
The triangle rate score (PYC file, result)
Our upcast sets
S.cerevisiae (triangles and lines), Eukaryotes (triangles and lines, Prokaryotes (triangles and lines), All interactions (triangles and lines)
Our results
The triangle rate score for 87,181 eligible protein pairs (fully annotated)
List of the high scoring protein pairs which are not in DIP
List of the low scoring protein pairs which are in DIP
The triangle rate score for 224,631 eligible protein pairs (fully and partially annotated)