About the full predicted scores (Yeast) downloading

 

 

        We made predictions on all possible Yeast pairs ( 6270 * 6270 ) for co-complex PPI task and the physical PPI task.

 

        We share these full predicted PPI score files, with each of them having19653315 lines

 

        Note, we have to point out that the positive training data used here were collected in 2004. They were small and noisy. There are quite some bigger and better Yeast PPI reference sets coming out recently. The predictions would be definitely improved with a better and larger positive training set. However our shared score files are still good baseline sets.

 

        Note: the following full predicted score files were derived from a modified feature version with 158 features.

o       Derived From our detailed 162 features version with the four homology based PPI features removed.

o       This is to avoid the double usage of the homology based PPI information, when using these computational scores for helping other organismsí PPI estimation.

o       In this way, our predictions were purely from yeast and thus, we could easily use the predicted evidence to derive PPI evidence for other organisms, like human and fly

 

 

        There are five files shared totally ( All are zipped for size limit)

o       Files Downloads directory:

o       Note: there is no self interacting

 

Protein Pairs List:

 

        File: full_pair_list.zip

        Containing the protein ORF pair list

        This list maps to all the other four score files

        Format in each line: ORF1ORF2

 

For Physical PPI task: SVM

        File: full_pair_list.dips.svm.result.zip

        Trained from Rand: 118184 ; Pos:2865

        SVM light (-t 1 -d 2 -j 15)

        Score range: [69.5920, -44.4740]

        Format:each line contains the score for a yeast protein pair

 

For Co-complex PPI task: SVM

        full_pair_list.mips.svm.result.zip

        Trained from : Rand: 236786 ; Pos:8236

        SVMlight para: (-t 1 -d 2 -j 15)

        Score range: [16.2720, -5.1995]

        Format:each line contains the score for a yeast protein pair

 

For Co-complex PPI task: RF

        full_pair_list.mips.rf.result.zip

        Trained from data: Rand: 236786 ; Pos:8236

        RF para: jtree = 300, jwt = 0.95

        Score range: [1, 0]

 

 

 

        We provided two PPI top cut list before. They were derived from SVM predictions for these two tasks. The scores got transformed to [0,1] in the old version of data share.

 

        For the files shared here, we put the raw predicted scores.

 

        Note: these three full predicted files were derived from thefeature version with 158 features.

o       Derived From our detailed 162 features version with the four homology based PPI features removed.

o       This is to avoid the double usage of the homology based PPI information, when using these computational scores for helping other organismsí PPI estimation.

o       In this way, our predictions were purely from yeast and thus, we could easily use the predicted evidence to derive PPI evidence for other organisms, like human and fly

 

o       FAQ page