- Help -

Step 1: Enter a protein sequence in the text box or upload a protein sequence file. The sequence(s) must be in FASTA format. When doing predictions for more than one sequence (maximum = 10 currently), each sequence must have the '>identifier' line required in FASTA format. N.B. Any sequences not separated by a '>' line will be treated as one sequence!

Step 2: (Option) Please enter your email address, where your prediction results will be sent. We encourage you to enter it, because the prediction for one sequence will usually take 10-20 minutes, but sometimes longer, depending on the sequence lengths and the server load.

Step 3: Click 'submit' to submit your protein sequence(s) for prediction.

 


After your submission, a page reporting job ID(s) and the status of your job(s) will appear. This page will automatically refresh every 30 seconds and update the job status. As soon as the results become available, the job status will change to "Available" and will also be reported by a change in colour of the folder icon . To access the results page, please click the folder icon. If you do not provide your email address to get the results by email, we encourage you to bookmark the web link of the results page so you are able to access it again at a later time.

  The job status will change to "Available".

 

 


Example Results for PSIVER

Your results will be sent within 10-20 minutes, but sometimes longer, depending on the sequence lengths or the server load. After receiving the email, please click on the URL(s) in the email and access your results page. N.B. The URL(s) will be automatically deleted after 10 days!

On the results page, the residues predicted to be involved in protein-protein interactions are shown as red-coloured residues in your query sequence. If you want more information on the predictions for the sequence, please view the graph or download the data file.

 


  A graph of the prediciton results can be viewed, please click "View Graph".

On the graph, the x-axis shows your query sequence and the y-axis shows the threshold for the prediction. More detail can be seen on the graph by zooming in; please click and drag on the plot to zoom (there will be a momentary delay while the plot redraws). To reset the plot, please double click on the plot or click the eReset Zoomf button.

The optimal threshold value (0.37), which can achieve the best performance (a Matthews Correlation Coefficient (MCC) of 0.151, a Precision of 30.6% and a Recall of 41.6%) in a leave-out one cross-validation on a dataset of 186 non-redundant protein seqeunces from the PDB, is rescaled to zero.

If predictions with a lower false positive rate are required, only the residues with higher threshold values should be considered. On the other hand, if predictions with a lower false negative rate are required, the residues with lower threshold values should also be considered.

To highlight only the residues above a threshold, please click and drag on the plot above that threshold. This graph function enables users to quickly determine which residues should be first considered for laboratory experimental verification.

  Click and drag on the plot to zoom

 


  A text format file (.pred) of the prediction results can be downloaded, please click "Download"'.

Example of the text format file:
#
# Prediction Result by PS-PPI @ NIBIO
# Mon Jan 25 16:50:05 JST 2010
#
# Threshold(MCC) = 0.370
# Total = 122
# Positives = 36 MCC:( 0.295)
# Negatives = 86 MCC:( 0.705)
# Average = 0.297
# Variance = 0.033
# Standerd deviation = 0.182
#
# PRED num class(MCC) aa dec z-value
#
PRED 1 - M 0.214 -0.453
PRED 2 - S 0.305 0.046
PRED 3 + D 0.496 1.097
PRED 4 - L 0.107 -1.040
PRED 5 - V 0.062 -1.290
PRED 6 - T 0.160 -0.750
PRED 7 + K 0.481 1.015
PRED 8 - F 0.084 -1.165 : :
PRED    121 - H  0.335  0.209 
PRED    122 + D  0.475  0.977 
#
#
   0 :  MSDLVTKFES LIISKYPVSF TKEQSAQAAQ WESVLKSGQI QPHLDQLNLV LRDNTFIVST
MCC: :  --+---+--- -----+--+- ------+-++ --+------- ---------+ ---+-+---+
  60 :  LYPTSTDVHV FEVALPLIKD LVASSKDVKS TYTTYRHILR WIDYMQNLLE VSSTDKLEIN
MCC: :  +++++++-+- --+--++-+- ---------- ---+-++--+ --++-+++-+ ---------+
 120 :  HD
MCC: :  -+

# END

The text file of the prediction results contains six columns:

  • record name "PRED"
  • residue number
  • prediction;  + = interface, - = non-interface
  • residue name (1-letter code)
  • score
  • z-value of scores



© PSIVER Copyright
PSIVER is maintained by Yoichi MURAKAM @ Bioinformatics Project, NIBIO