AroundA is a classifier, AroundA model finetune from InstaDeepAI/nucleotide-transformer-v2-500m-multi-species model. AroundA predicts the likelihood of a nucleotide site of interest to be modified based on the upstream and downstream sequences of its RNA.
This model was trained based on the A (random A bases on exons), A-to-I, m6A, Am, m6Am, m1A sites, using 50nt sequences upstream and downstream of the site of interest as context.
Training and test set: for non-Chr8, Chr14 data, random split. 80% for training set, 20% for test set.
evaluation_metrics
Evaluation set: Chr8, Chr14 data
AUC
PRAUC
Confusion
In the figure class0 is A,class1 is A-to-I, class2 is m6A,class3 is Am, class4 is m6Am and class5 is m1A.
