Predição de interações entre piRNAs e elementos transponíveis por meio de predictive bi-clustering trees

• Autor(a) principal: Freire-Oliveira, Hiago
• Data de Publicação: 2022
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/16064
• topic: piRNA; Elemento transponível; Predição de interação; Árvore de decisão; Aprendizado de máquina; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO::LINGUAGENS DE PROGRAMACAO

PIWI-interacting RNAs (PiRNAs) are a class of interfering RNAs whose actions range from regulating gene expression to fighting viral infections and silencing transposable elements, possessing unique characteristics such as being from 21 to 35 nucleotides long, displaying an uracil bias at the end 5' and 2'-O-methylation at the 3' end. Transposable elements (TEs) are genetic elements capable of moving between host genomes, being split into retrotransposons and DNA transposons. The replication of TEs can promote harmful recombination events by generating breaks in DNA double strands, in addition to interference in expression, considering that their promoters can lead to aberrant transcription of neighboring genes. Silencing of these elements by piRNAs occurs in the germ line in the majority of animals and is essential for the maintenance of genome integrity. The problem of in silico prediction of interaction between piRNAs and TEs was addressed by machine learning using a decision tree algorithm, namely Predictive Bi-Clustering Trees (PBCT). In order to improve the algorithm’s performance, the interaction matrix of piRNAs and TEs was reconstructed by means of an Beta-distribution-rescored Neighborhood Regularized Logistic Matrix Factorization (NRLMFβ) algorithm. PBCT was applied in 5-fold and 10-fold cross-validation configurations, both for the matrix without reconstruction (BICT) and for the matrix reconstructed by NRLMFβ (BICTR). In general, PBCT applied to this dataset was not able to predict positive interactions satisfactorily, behaving as a random classifier. Comparatively, in the BICT method, PBCT presented higher values of AUROC and AUPRC. However, in the BICTR method, PBCT was able to correctly predict more positive interactions, which are, in fact, the major interest in this study. Potential biological applications and ways to improve the algorithm’s performance were also considered.