The University of Kansas Proteomics Service (KUPS) provides high-quality protein-protein interaction (PPI) datasets for researchers who are interested in eulicidating PPIs with 'in silico' methods.
KUPS now combines three manually curated PPI databases (i.e, IntAct, MINT and HPRD ) with features and annotations retrieved from UniProt , Position Specific Scoring Matrix (PSSM) with non-redundent database from NCBI PSI-BLAST , and amino acid scaling features from the Amino Acid Index database .
Positive training examples can be filtered by their host species, interaction type (e.g. targetting only antigen-antibody interactions), experimental derivation (e.g. filter out interactions results from two-hybrid) (see how to use it).
Negative training examples can be generated by four different methods: uniform random pairs, functionally dissimilar pair, spatially separate pairs and non-interacting domain pairs. (see their definitions and usages).
KUPS provides serveral features that can be used for creating train/test data such as protein sequences, Position Specific Score Matrix (PSSM), physicochemical/biochemical scales from AAindex and so on(see how to use it).
This work was supported by National Science Foundation CAREER Award IIS-0644366 and the associated REU.