Mining transcriptome data to identify genes and pathways related to lemon taste using supervised and unsupervised data learning methods

Abstract

There is a dearth of studies on the genes engaged in citrus taste. Unraveling the major genes involved in pathways related to the taste of citrus (sweet or acidic) is highly important for developing new genotypes with favorable taste. Pivotal genes linked to citrus taste can be extracted through mining a large number of expression data. To attain this objective, 10 different attribute weighting algorithms (AWAs) were applied on the expression data from three lemon (Citrus limon) genotypes differing in terms of fruit acidity. As a result, a total of 170 probe sets were identified by more than eight AWAs as the most discriminative probe sets. Subsequently, principal component analysis and hierarchical clustering heatmaps were implemented for validation of the 170 top-ranked probe sets. Noticeably, the identified top 170 probe sets significantly contributed to accurate discrimination between sweet and acidic lemon samples, which indicate the significance and accuracy of prediction of probe sets. According to the results, some genes like citrate synthase, malate dehydrogenase, proton-pumping ATPase, and flavanone 3-hydroxylase had distinct roles in differentiation of the studied genotypes and acidity. Among all of the genes, malate dehydrogenase was the most informative. To the best of our knowledge, this is the first report on identifying the most important genes contributing to lemon taste using supervised and unsupervised data learning methods.