Premium
HERMeS: A second generation approach to the automatic analysis of two‐dimensional electrophoresis gels. Part V: Data analysis
Author(s) -
Tarroux Philippe,
Vincens Pierre,
Rabilloud Thierry
Publication year - 1987
Publication title -
electrophoresis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.666
H-Index - 158
eISSN - 1522-2683
pISSN - 0173-0835
DOI - 10.1002/elps.1150080404
Subject(s) - computer science , principal component analysis , cluster analysis , pattern recognition (psychology) , artificial intelligence , two dimensional gel electrophoresis , multivariate statistics , expression (computer science) , set (abstract data type) , data mining , field (mathematics) , biological system , machine learning , chemistry , mathematics , biology , biochemistry , pure mathematics , proteomics , gene , programming language
Data analysis programs are essential tools for processing the large quantity of data produced by two‐dimensional gel electrophoresis computerized analysis. Global handling of two‐dimensional gel information can be achieved using both analysis programs and artificial intelligence techniques. First, a statistical multivariate approach is presented, used to sort gels and to obtain a gel classification according to protein patterns. Both clustering and multivariate techniques are used and the results show that a classification of protein expression patterns according to effector action can be achieved. Spots can also be sorted and compared according to their expression under effector action. Each spot is thus described using a set of variables (its intensity for each gel pattern) and its description is essentially given in a multivariate way. Therefore, each facet of the regulation of a particular protein according to the action of several effectors can easily be taken into account. Spot sets containing spots having common regulation characteristics are obtained using principal component analysis. We show that each group corresponds to a functional aspect of the regulation in a given cell line. The map obtained describes the organization of cell functions and the interactions of these functions. Another approach employs artificial intelligence techniques. Since the rules available for describing some aspects of knowledge in the field of two‐dimensional electrophoresis remain unknown, learning techniques are suitable for defining them. Their use for the automatic recognition of spot groups belonging to the same entity on a gel is illustrated. Our learning program is applied to the determination of glycosylation groups and to the identification of relevant spot patterns.