A novel and effective way of approaching high class-imbalance in pre-miRNA prediction. The unsupervised deep model can overcome the problem of having very few positive class labels.
In this webdemo, two models were previously trained for plants and animals with data from miRBase v17.
Extracted features of positive samples from miRBase v18-19 are provided for testing.
- Animals: {Bombyx mori} {Caenorhabditis elegans} {Ciona intestinalis} {Homo sapiens} {Macaca mulatta} {Mus musculus} {Oryzias latipes} {Pongo pygmaeus} {Rattus norvegicus} {Taeniopygia guttata} {Tribolium castaneum}
- Plants: {Arabidopsis thaliana}{Cucumis melo} {Glycine max} {Hordeum vulgare} {Malus domestica} {Medicago truncatula} {Nicotiana tabacum} {Oryza sativa} {Populus trichocarpa} {Sorghum bicolor}
In the output maps: miRNA neurons in red, no-miRNA neurons in blue, and input sequences in black.
Contact: Georgina Stegmayer