RBPCNN

RBPCNN predicts sequence specificities of RNA-binding proteins by applying a convolutional neural network to raw RNA sequences and conservation scores to infer binding motifs and interaction preferences.

Key Features:

• Convolutional neural network: Employs a CNN architecture tailored to integrate raw RNA sequences with conservation scores to capture patterns governing RBP–RNA interactions.
• Evolutionary information: Incorporates conservation scores as an additional input feature to represent evolutionary constraints on RNA sequences relevant to binding specificity.
• Motif extraction: Automatically extracts sequence motifs associated with RBP binding from learned CNN representations.

Scientific Applications:

• RBP specificity mapping: Improves prediction of RBP–RNA interactions to support mapping of binding specificities across transcripts and genomes.
• Post-transcriptional regulation studies: Supports investigation of regulatory mechanisms mediated by RBPs in post-transcriptional gene regulation.
• RNA element annotation: Facilitates identification and annotation of sequence motifs and RNA elements implicated in RBP binding and discovery of regulatory pathways.

Methodology:

Trains a convolutional neural network on datasets containing raw RNA sequences and conservation scores, learns from primary sequence and evolutionary context, and is evaluated against state-of-the-art methods reporting a 2.67% increase in AUC (area under the receiver operating characteristic curve) and an 8.03% increase in mean average precision.