Predicting microrna expression from sequence

The improvements in technology has drastically accelerated the scientific activities in all domains. The increase in the number of studies has allowed researchers to reuse existing data obtained from previous experiments. In this context, working with complete and perfect data has become an urgent need. Therefore, working with the perfect data has become an important condition. In last years, scientists have proven that, cancer is related with the activities of microRNAs. Therefore, the experiments with microRNA have increased dramatically. A huge amount of data has been inserted to related databases such as Gene Expression Omnibus. Scientist measured gene expression of microRNA using microarray technology however, some of the data measured incorrectly or it could not be measured. Most of studies had to wait for gene expression evaluation of recently discovered miRNA's. In this study, we try to predict missing data of the given promoter sequence of a micro RNA. We attempt to predict its expression using regression models ( linear regression, KNN regression and RVM regresssion ) learned from the expression levels of other microRNAs obtained through a microarray experiment. To our knowledge, this is the first study that evaluates the predictability of microRNA expression from sequence. RVM regression which uses Gaussian kernel gave us the most successful results. The results encourage the use of the system for microarray missing data imputation or completing old experiments with new explorations.