Classification of Oncologic Data with Genetic Programming

摘要

Discovering the models explaining the hidden relationship between genetic material and tumor pathologies is one of the mostimportant open challenges in biology and medicine. Given the large amount of data made available by the DNA Microarraytechnique, Machine Learning is becoming a popular tool for this kind of investigations. In the last few years, we have beenparticularly involved in the study of Genetic Programming for mining large sets of biomedical data. In this paper, we presenta comparison between four variants of Genetic Programming for the classification of two different oncologic datasets: the first onecontains data from healthy colon tissues and colon tissues affected by cancer; the second one contains data from patients affected bytwo kinds of leukemia (acute myeloid leukemia and acute lymphoblastic leukemia). We report experimental results obtained usingtwo different fitness criteria: the receiver operating characteristic and the percentage of correctly classified instances. These results,and their comparison with the ones obtained by three nonevolutionary Machine Learning methods (Support Vector Machines,MultiBoosting, and Random Forests) on the same data, seem to hint that Genetic Programming is a promising technique for thiskind of classification.