SINC(I)   25518
INSTITUTO DE INVESTIGACION EN SEÑALES, SISTEMAS E INTELIGENCIA COMPUTACIONAL
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
Deep neural architectures for highly imbalanced data in bioinformatics
Autor/es:
BUGNON, LEANDRO; MILONE, DIEGO; YONES, CRISTIAN; STEGMAYER, GEORGINA
Lugar:
Salta
Reunión:
Simposio; AGRANDA; 2019
Institución organizadora:
Sociedad Argentina de Informática e Investigación Operativa (SADIO)
Resumen:
The classification task with imbalanced data has been largely recognized as an important issue in machine learning [3, 4, 8]. Most machine learning algorithms work well with balanced datasets, but with imbalanced datasets supervised classifiers tend to be biased towards the majority class and have a very low performance on the minority one. This is of particular importance in bioinformatics, including studies such as diagnosis based on gene expression data, protein function classification, activity prediction of drug molecules and recognition of precursor microRNAs (pre-miRNAs) [5]. MiRNAs are a special type of non-coding small length RNA, which can be critical regulators in the gene expression. They may determine the genetic expression of cells and influence the state of the tissues. In a real-life scenario, the number of known pre-miRNAs is usually very small in comparison with other unknown sequences, reaching imbalances of 1:1,000 or higher.