|
SUMMARY: UCSD researchers have developed a novel algorithm for the prediction of the positive interactions of proteins with other proteins, nucleic acids, small molecules and biopolymers. The system is first trained to recognize patterns that characterize positive interactions of proteins within a proteome, based on primary protein sequences and associated physiochemical information available in databases. Known molecular pathways and structural information relating to chemical interactions are then used to make predictions of the abilities of individual proteins to interact with other molecules in a pair-wise fashion (Whole-proteome interaction mining). The accuracy of these predictions in yeast is over 80%.
Recently, the method has been substantially improved such that the algorithm can make these predictions across proteomes, with a success rate higher than is available with other methods. This characteristic enables one to make protein interaction predictions for proteomes for which little experimental data are available.
POTENTIAL COMMERCIAL APPLICATIONS: This software is useful for virtual screening and validation of drug candidates that target proteins or protein/DNA interactions.
One can also use the algorithm to predict protein epitopes of functional domains, and binding sites for designer antibodies and biocompatible polymers, as well as a tool for basic research, such as for cell signaling pathways.
This technology is available for licensing or sponsored research. More information can be obtained under a Confidentiality Agreement.
CASE NUMBER: SD2002-033
KEY WORDS: Drug discovery, Protein interactions.
INQUIRIES TO: invent@ucsd.edu
|
