Functional annotation of unknown proteins is am ajor goal in proteomics. Ak ey step in this annotation process is th ed efinition of ap rotein' ss ubcellular lo- calization. As ac onsequence, numerous predictio nt echniques for localization have been developed ove rt he years. These methods typically focus on as ingle underlying biological aspect or predict as ubset of all possible subcellula rl ocalizations. There is ac lear need for ne wm ethods that utilize an dr epresent available protein specific bio- logica lk nowledge from several sources, in order to improve accurac ya nd localization coverage for aw ide range of organisms. Her ew ep resent an ove lS upport Vector Machine (SVM)-based approach for pre- dictin gp rotein subcellular localization, which integratesinformatio na bout N-terminal targeting sequences ,a mino acid composition, and protein sequence motifs. An impor- tant step is taken towardsemulating th ep roteinsorting process by capturingand bring- ing together biologically relevant information. Our nove la pproach has been used to develop tw on ew prediction methods, TargetLoc and MultiLoc. TargetLoc is restricted to analysis of proteins containing N-terminal targeting sequences, whereas MultiLoc covers all major eukaryotic subcellular localizations for animal, plant, an df ungal pro- teins. Compared to simila rm ethods, TargetLo cp erforms better than these. MultiLoc performs considerably better than comparable prediction methods predicting al lm ajor eukaryotic subcellula rl ocalizations, an ds hows better or comparable results to meth- ods tha ta re specialized on fewer localizations or for one organism.