RNA etkileşimlerinin in silico analizi

Many supervised machine learning models have been developed for the classification and identification of non-coding RNA (ncRNA) sequences. These models play a significant role in the diagnosis and treatment of various diseases. During such analyses, positive learning datasets typically consist of known ncRNA examples, some of which may even be confirmed with strong experimental evidence. However, there is no database of validated negative sequences for ncRNA classes or standardized methodologies for generating high quality negative samples. To overcome this challenge, a new method for generating negative data called the NeRNA (Negative RNA) method has been developed in this study. NeRNA generates negative sequences using known ncRNA sequences and their octal representations, similar with frame shift mutations found in biology but without base deletions or insertions. In this thesis, the NeRNA method was tested separately with four different ncRNA datasets, including microRNA (miRNA), transfer RNA (tRNA), long non-coding RNA (lncRNA), and circular RNA (circRNA). Additionally, a species-specific case study was conducted to demonstrate and compare the performance of the study's miRNA predictions. The results of 1000-fold cross-validation on machine learning algorithms such as Decision Trees, Naive Bayes, Random Forest classifiers, and deep learning algorithms like Multilayer Perceptrons, Convolutional Neural Networks, and Simple Feedforward Neural Networks showed that models developed using datasets generated by NeRNA exhibited significantly high prediction performance. NeRNA has been published as an easy-to-use, updatable, and modifiable KNIME workflow, along with example datasets and required extensions that can be downloaded and utilized. NeRNA is designed specifically as a powerful tool for RNA sequence data analysis.