ResNet, short for Residual Network, stands as a groundbreaking convolutional neural network (CNN) architecture introduced in 2015 by Kaiming He at Microsoft Research. Its pivotal innovation lies in the incorporation of residual connections, known as "skip connections," which tackle the challenge of training exceptionally deep networks. These connections enable the network to learn residual functions, mitigating the vanishing gradient problem encountered in deep architectures. Within each residual block, multiple convolutional layers are followed by a shortcut connection that adds the original input to the output of the convolutional layers. This unique design empowers ResNet to effectively learn complex mappings while maintaining ease of optimization, facilitating the training of networks with hundreds or even thousands of layers. As a result, ResNet has emerged as a cornerstone in computer vision, consistently achieving state-of-the-art performance across various tasks such as image classification, object detection, and image segmentation.

Ensemble modeling stands as a powerful technique in machine learning, where multiple diverse models are combined to improve predictive performance over individual models. By leveraging the wisdom of the crowd, ensemble models harness the complementary strengths of different algorithms to achieve superior accuracy, robustness, and generalization. These models can range from simple methods like averaging predictions to more sophisticated techniques such as bagging, boosting, and stacking. Ensemble methods mitigate the risk of overfitting and increase model stability by reducing variance and bias, respectively. They find widespread application across various domains, including classification, regression, and anomaly detection, consistently delivering state-of-the-art results in machine learning competitions and real-world scenarios.

Users can input or upload the query sequences in the txt format. The txt file should include five columns which are seqnames, start, end, width, and strand respectively. We kept two modes "Resf5C Only" and "Ensemble Model (Resf5C + XGB)" for users to select different modes to use. Just click the corresponding option in the button under "modes". Results will be presented or downloaded after a while.

The result table shows chromosome, position, strand, probability of the potential f⁵C sites, likelihood ratio with its corresponding confidence level and so on.

We displayed a framework figure and a brief introduction of the model (Please refer to "Model" page). The model's parameter can be downloaded in the model page and read as a pth file.