Genetic algorithms for deciphering the complex chemosensory code of social insects

Chemical communication among social insects is often studied with chromatographic methods. The data generated in such studies may be complex and require pattern recognition techniques for interpretation. Presently, we are analyzing gas chromatographic (GC) profiles of hydrocarbon extracts obtained from the cuticle and postpharyngeal gland (PPG) of 400 Cataglyphis niger ants using a genetic algorithm (GA) for pattern recognition analysis to identify the factors influencing colony odor. The pattern recognition GA identifies features (i.e., chromatographic peaks) whose principal component plots show clustering of the samples on the basis of class. Because the largest principal components capture the bulk of the variance in the data, the peaks chosen by the GA primarily convey information about differences between the classes in a data set. As it trains, the pattern recognition GA focuses on those classes and/or samples that are difficult to classify by boosting their class and sample weights. Samples or classes that consistently classify correctly are not as heavily weighted as samples or classes that are difficult to classify. Over time, the algorithm learns its parameters in a manner similar to a neural network. The proposed algorithm integrates aspects of artificial intelligence and evolutionary computations to yield a “smart” one-pass procedure for feature selection and pattern recognition. Utilizing the pattern recognition GA, two specific questions were addressed in this study: (1) Does the overall hydrocarbon profile of the colony change with time? (2) Does the queen influence the hydrocarbon pattern of the colony?