With a background in Computer Science and solid experience in machine learning (ML) and non-coding RNAs (ncRNAs), Mirele conducted research on the performance of ML methods for ncRNA prediction using in silico-generated data. During her doctorate, she is deepening this line of research with a comparative study of the performance of transfer learning (TL) models against BLAST in ncRNAs prediction, incorporating taxonomic distances into the analysis.

During her stay in Leipzig, she conducted experiments to quantitatively compare homology-based search methods (such as BLAST) and ML methods (such as graph convolutional networks) with and without TL for classifying ncRNA sequences with flanking regions. The experiments used three datasets with low, medium, and high taxonomic distances. The study also investigated how these regions affect model performance since identifying ncRNAs in a genomic context where unknown regions flank the sequence of interest is substantially more complex.

Mirele states: “SECAI's support was essential for this experience to take place; without the scholarship, it would not have been possible to carry out this period of research in Germany. This opportunity transformed my journey as a scientist and broadened my perspective as a researcher.” Mirele's stay at SECAI represents a significant step in consolidating scientific cooperation between Brazil and Germany in the field of Bioinformatics and AI. The exchange not only strengthens academic cooperation but also contributes to the training of skilled researchers.