Fig 1.
Positive and negative datasets were obtained (Step 1). Features of the datasets were extracted to obtain 366-dimensional vectors for each site as training data. The SVM classifier was trained to generate the SVM model and the performance of the model was evaluated (Step 2). Human transcriptome-wide m6A sites were predicted and a web server was constructed (Step 3).
Fig 2.
Overall Performances of Mammalian Classifiers Based on 5-fold Cross-validation Tests.
(A) The ROC curve illustrating the performance for full transcript mode. (B) The ROC curve illustrating the performance for mature mRNA mode.
Fig 3.
Performances of the Mammalian Classifiers on Independent Testing Datasets.
(A) ROC curve illustrating the performance on the unbalanced independent testing dataset in full transcript mode. (B) Precision-recall curve illustrating the performance on the unbalanced independent testing dataset of full transcript mode. (C) ROC curve illustrating the performance on the unbalanced independent testing dataset of mature mRNA mode. (D) Precision-recall curve illustrating the performance on the unbalanced independent testing dataset of mature mRNA mode.
Table 1.
Performance of RNAMethPre for various stringency thresholds and comparison with SRAMP.
Table 2.
Comparison of RNAMethPre with the Existing Web Server SRAMP using Independent Unbalanced Datasets.
Fig 4.
The user interface of the RNAMethPre web server.
Fig 5.
The genome browser to visualize the query results.