- A graph of the result has only one dot in the center.
- Is multiple sequence alignment absolutely required?
- Can't the server perform multiple sequence alignment?
- How does the program define surface residues?
- Why is ID necessary?
- Is there a text output for the score?
- Why is the prediction slow?
- A non-RNA binding protein is predicted to have RNA interface residues. Why?
- What is the detail of the prediction method?
- How can I access the tool from the unix command line?
A graph of the result has only one dot in the center.
There are a number of possibilities:
- The input file contains more than one chain. At the moment multiple chain is not acceptable. Please upload a PDB file containing a single chain.
- A sequence with ID does not exist in the alignment file.
- A amino acid sequence with ID in the alignment is different from the one of PDB. Be sure that selenomethionine is changed to methionine in PDB, for instance.
Is multiple sequence alignment absolutely required?
A multiple sequence alignment is required when the prediction needs a
profile. When you use [S], [AS], [ASD], or [A2SD], in other
words the methods without P, an alignment is not needed.
Can't the server perform multiple sequence alignment?
Actually yes. Follow the link here. But
our server machine is not powerful enough to perform homology search
and a huge multiple sequence alignment, sorry. We are think about updating
the server in the future.
How does the program define surface residues?
Accessibility of every residue is calculated based on the input PDB file,
and all the residue with accessibility more than 0.08 is considered as
surface residues.
Why is ID necessary?
The prediction program is not sophisticated enough to tell which of the
sequence in the multiple sequence alignment is identical to the amino
acid sequence in PDB. We will update the program shortly.
Is there a text output for the score?
At the moment no. Please download the prediction result file in PDB
format and select a value shown in the column of temperature factor
of ATOM row. By selecting the values of CA only, one can obtain the
score.
Why is the prediction slow?
A number of reasons. Slow network traffic, slow programs, slow server.
Just be patient, svp.
A non-RNA binding protein is predicted to have RNA interface residues. Why?
The question that the prediction method is trying to answer is "Here is a
RNA-binding protein. Where is the RNA binding surface?" Therefore, if one
perform prediction on non-RNA-binding proteins, the usage is out of the
range of the question that we tried to answer by the method. The question
such as "Here is a protein. Does this protein bind to RNA?" is, however,
an important question and we would try to have a method to answer this
type of question in the future. Besides, there already are a couple of
methods to answer this type of question such as BIDD.
What is the detail of the prediction method?
Read help page. Further detail of the method is now
under review at Nucleic Acids Research Editorial Board.
How can I access the tool from the unix command line?
Here is a C++ program to use KYG method from a unix
command line. With curl, a unix command, you can use this website from
your program.
curl -F "method_type=8" -F pdb_file=@1ETF_P.pdb http://cib.cf.ocha.ac.jp/KYG/KYG.php | ./parseKYG.exe
This tool is kindly provided by Dr. Marc Parisien at University of Montreal, Canada (Merci!).