Modelling the relationship between the microarray data of the nci-60 anticancer dataset with therapeutic responses by genetic programming
Leonardo Vanneschi, Ilaria Giordani, Elisabetta Fersini · BOA (University of Milano-Bicocca) · 2007
Predicting in an accurate way drug response of individual patients is an important task for personalized medicine. In the last few years, pharmacogenomics research in chemosensitivity prediction has presented a large number of contributions aimed at finding the correlation between gene and drug based on transcriptional profiling. However, proteomic profiling will more directly solve the current functional and pharmacologic problems. In this paper, we try to establish a mathematical relationship between genomic features and therapeutic responses to a particular pharmacologic treatment by means of Genetic Programming, using the NCI-60 Anticancer Dataset. In order to improve Genetic Programming generalization ability, training has been performed by means of a dynamic use of validation sets and functions have been evaluated using root mean squared error and correlation coefficient between outputs and targets. Experimental results are promising and should pave the way to a larger experimentation of Genetic Programming for predicting cancer drug response.