SubNucPred is a combined method of two different prediction approaches which predicts 10 different sub-nuclear locations namely: centromere, chromosome, nucleolus, nuclear envelope, nuclear speckles, telomere, nucleoplasm, nuclear matrix, PML body and nuclear pore complex. When a protein is submitted to SubNucPred for prediction, first it searches for presence of Pfam domain and matches it with our curated list of sub-nuclear location-specific exclusive domains. The list contains only domains which are found in a single sub-nuclear location, nowhere else. If the query protein have atleast one Pfam domain which is common to our curated domain list, the query protein is assigned to the location according to the domain. In case it could not find any exclusive domain, the algorithm forward the query protein to the 2nd method which does prediction using amino acid composition based SVM module.
The overall prediction accuracy of SubNucPred is 85.05% for centromere, 76.85% for chromosome, 81.79% for nucleolus, 81.27% for nuclear speckle, 79.37% for nuclear envelope, 83.33% for telomere, 88.89% for nuclear pore complex, 76.98% for nucleoplasm, 77.78% for nuclear matrix, 75.40% for PML body respectively.
If you are using this webserver, please cite:
Kumar R, Jain S, Kumari B, Kumar M (2014) Protein Sub-Nuclear Localization Prediction Using SVM and Pfam Domain Information. PLoS ONE 9(6): e98345. doi:10.1371/journal.pone.0098345