Machine Learning for Prediction of Protein-Protein Interactions

Although the ability of one protein to bind to another one is determined by the three dimensional structure [1] there are several approaches who use only the primary structure of the proteins and some additional information to detect protein functions and protein-protein interactions [2], [3]. In [4] it is proposed that interacting proteins co-evolve and that the amount of co-evolution of two proteins can be accessed by an examination of the corresponding phylogenetic trees.
The master thesis deals with the improvement of the mirrortree method [4] by means of dimensionality reduction and transductive learning. Initially the phylogenetic distances are directly used to build a high dimensional space. Dimensionality reduction methods like PCA are applied and the distances between interacting proteins in the reduced space are examined. Afterwards the phylogenetic distances are used to build a kernel matrix with the correlations between every possible protein pair of the data set and some new kernel based learning algorithms are compared. The results suggest that it is possible to improve the mirrortree method significantly.

[1] Jones,S. and Thornton,J.M. (1996) Principles of protein–protein interactions. Proc. Natl Acad. Sci. USA, 93, 13–20.
[2] Pellegrini M, Marcotte EM, Thompson MJ, Eisenberg D, Yeates TO: Assigning protein functions by comparative genome analysis: protein phylogenetic profiles. Proc Natl Acad Sci USA 1999, 96:4285-4288.
[3] Marcotte EM, Pellegrini M, Ho-Leung N, Rice DW, Yeates TO, Eisenberg D: Detecting protein function and protein-protein interactions from genome sequences. Science 1999, 285:751-753.
[4] Pazos F, Valencia A: Similarity of phylogenetic trees as indicator of protein-protein interaction. Protein Eng 2001, 14:609-614.