NpPred is a method developed for predicting nuclear proteins. This method has been developed on a non-redundant dataset consists of 2710 nuclear and 7662 non-nuclear proteins. During development of NpPred we developed number of SVM based methods using various types of composition (amino acid, dipeptide, split) and achieved maximum accuracy 85.47% when evaluated using five fold cross-validation. Using hybrid approach (Pfam domain and SVM) accuracy increased to 94.61%. This method performed better than existing methods when evaluated on independent dataset obtained from BaCelLo (Pierleoni et al., 2006) and NucPred (Brameier et al., 2007). In this server we have given 2 approaches for prediction (a) SVM module developed using N-terminal 25 and remaining residues amino acid composition and (b) Hybrid approach combining SVM module and HMM profile. We hope this method will be useful for researcher working on field of genome annotation.
If you are using this web-server please cite: Kumar, M. and Raghava, G.P.S. Prediction of Nuclear Proteins using SVM and HMM Models. BMC Bioinformatics. 2009 Jan 19;10(1):22