AdaBoost Algorithm with Random Forests for Plant and Animal Precursor MicroRNAs Classification

Recently, there have been many reports about miRNAs which have played important regulatory roles. There are differences in control mechanism pathways between plant and animal miRNAs. It is hypothesized that these differences may originate from differences in the characteristics of precursor miRNA structures between animals and plants. This work proposes a combination of AdaBoost and random forests in order to discriminate between plant and animal precursor miRNAs. Many features related to the precursor miRNAs characteristics from many literatures were collected and filtered. Seven classification models were compared, including naïve bayes, neural network, k-nearest neighbor, decision tree, support vector machine, random forests and Adaboost random forests. Based on independent test sets, the results showed that the combination of the Adaboost and the random forests model achieved the highest accuracy of 85% in discriminating animals and plants.