Overveiw
CalPred is a "tool for EF-hand calcium binding protein
prediction and calcium binding region identification" using machine learning
techniques. It is a free web based software package and is accessible via world
wide web from various platforms. It integrates a range of currently available
open source and / or free software packages such as SNNS, EMBOSS and SVMlight,
Cygwin and Microweb server for the purpose of anlysing the protein nature and
identification of the calcium binding regions in the given protein. CalPred
Server is available for free download as a portable server to promote open source
spirit and to reduce the load on our servers.
CalPred
key features: Usability of program.
The key features of CalPred include:
- It gives output of protein statistics from PEPSTATS program of EMBOSS
package that has been used to train the "first level" filter of
the CalPred program.
- It detects sequences which belong to the calcium binding protein family,
but have not been picked up by the "EF-hand calcium-binding domain"
pattern sited as PS00018 entry of Prosite database.
- It predicts specific calcium-binding regions in the query protein, i.e.
the prediction is done for every amino acid residue present in the protein.
- It is free of charge.
- It is downloadable as a open source project in form of standalone portable
server for in-house use.
What
are the different applications and how are they organised?
Currenty there are ten applications incooperated in CalPred. These
different applications are:
-
ANNpepstats :
It takes protein properties from PEPSTATS module of EMBOSS package as input
and queries a neural network model of architecture (51-4-1) to predict the
nature of protein i.e. whether its calcium binding or not.
-
ANNbinary : It takes
a protein sequence as input and encodes it in binary format (for more info.
see user docs) and queries a neural network model of architecture (260-20-1)
to predict the nature of a particular amino acid residue.
-
ANNpssm : It takes
a protein sequence as input and creates its PSSM matrix using PSI-BLAST
(for more info. see user docs) and queries a neural network model of architecture
(260-20-1) to predict the nature of a particular amino acid residue.
-
Prosite Scan : It takes a protein
sequence as input as performs simple pattern mathching using "EF-hand
calcium-binding domain" pattern i.e. PS00018 entry of Prosite database.
-
SVMpepstats_linear : It
takes protein properties from PEPSTATS module of EMBOSS package as input
and queries a support vector machine (SVM) model with a "linear"
kernel type to predict the nature of protein.
-
SVMpepstats_polynomial : It
takes protein properties from PEPSTATS module of EMBOSS package as input
and queries a support vector machine (SVM) model with a polynomial kernel
type to predict the nature of protein.
-
SVMpepstats_radial_bais : It
also takes protein properties from PEPSTATS module of EMBOSS package as
input to query a support vector machine (SVM) model with a "radial
bais" kernel type and predicts the nature of protein.
-
SVMpepstats_sigmoidal_tanh : It
also takes protein properties from PEPSTATS module of EMBOSS package as
input and queries a support vector machine (SVM) model with a "sigmoidal
tanh" kernel type and predicts the nature of protein i.e. whether its
calcium binding or not.
-
SVMbinary : It takes
a protein sequence as input and encodes it in binary format (for more
info. see user docs) and queries a support vector machine (SVM) model
with a "linear" type kernal to predict the nature of a particular
amino acid residue.
-
SVMpssm :
It takes a protein sequence as input and creates its PSSM
matrix using PSI-BLAST (for more info. see user docs) and queries a support
vector machine (SVM) model with a "linear" type kernal to predict
the nature of a particular amino acid residue.
These different applications are organised to form a workflow
as depicted in the figure below. To know more about the workflow and the validity
of the models see the User documentation.

CalPred
citation.
"CalPred: A tool for EF-hand calcium binding protein prediction
and calcium binding region identification."
Jaiswal, K.; Kumar, C.; and Naik, P. K.
Department of Bioinformatics and Biotechnology.
Jaypee University of Information Technology, India.
Disclaimer.
This software is free only for non-commercial use. It must not
be distributed as whole without prior permission of the author but some parts
of the software can be redistributed and / or modified as stated in the License.
The author is not responsible for implications from the use of this software.