Long terminal repeats are structural components of LTR retrotransposons, mobile (transposable) sequence elements in eukaryotic genomes. They are highly variable and repetitive regions of DNA. While transposable elements have been subject to research for a while, few tools to reliably classify and analyze their LTRs currently exist. The goal of this work is to apply state-of-the-art data analysis and machine learning algorithms to help detect and classify these elements. Learning data should be based on manually or semi-manually managed collections of proven LTR retrotransposons, such as GyDB, TREP, SOYBASE, InpactorDB that have already been classified into superfamilies or families, as well as general sequence data from public collections (NCBI, Phytozome). The work should focus on: - Reliable preprocessing of DNA sequences for use in deep learning tasks - Large data analysis and pattern mining algorithms (ECLAT) that have shown some promise in sequence analysis, from traditional ML models (k-NN, random forest, gradient boosting) to modern neural networks and transformers (RNN/LSTM, BERT /DNABERT) - Model interpretability techniques/DNABERT)