HYDRA WEB

We present a hybrid ensemble RBP classifier (HydRA), which leverages information from both intermolecular protein interactions and internal protein sequence patterns to predict RNA-binding capacity with unparalleled specificity and sensitivity using support vector machines (SVMs), convolutional neural networks (CNNs), and Transformer-based protein language models.

Occlusion mapping by HydRA robustly detects known RBDs and predicts hundreds of uncharacterized RNA-binding associated domains, accelerating construction of a comprehensive RBP catalog and expands the diversity of RNA-binding associated domains.

About:

HydRA has incorporated the predictions from HydRA-seq and SONAR3.0 via statistical integration and has higher overall sensitivity, specificity, and precision than HydRA-seq and SONAR3.0. But in practice, whether to combine/intersect the positive prediction results from HydRA, HydRA-seq, and SONAR3.0 or “HydRA only” depends on the task at hand. For instance, if our goal is to create a comprehensive and exhaustive list of RBPs, meaning that we only care about sensitivity, we can combine the positive predictions from all three classifiers. However, if we aim to create a highly reliable list of RBPs, prioritizing precision and specificity, we can take the intersection of the positive predictions from the three classifiers. Nevertheless, combining positive predictions will often decrease specificity and precision, whereas taking the intersection will negatively impact sensitivity. Thus, by default, we assume we need a list of predictions with high specificity, precision, and sensitivity simultaneously, which is achieved by using the HydRA-only strategy.

HydRA: Deep-learning models for predicting RNA-binding capacity from protein interaction association context and protein sequence.

About: