ac4C-AFL

ac4C-AFL precisely identifies human mRNA N4-acetylcytidine (ac4C) sites from primary RNA sequences using adaptive feature representation learning and ensemble feature selection.

Key Features:

• Adaptive Feature Representation: Extracts representative features from RNA sequences using an adaptive strategy with optimized sequence length.
• Ensemble Feature Importance Scoring: Ranks features using a novel ensemble strategy to identify the most relevant features for prediction.
• Sequential Forward Search: Determines the optimal feature set from 16 sequence-derived feature descriptors.
• Two-Step Feature Selection: Identifies the most efficient baseline models from 176 models built with 11 classifiers, then integrates their predicted scores for final model training.

Scientific Applications:

• ac4C Site Prediction: Genome-wide identification of N4-acetylcytidine modification sites in human mRNA to support studies of gene expression regulation.
• RNA Modification Research: Facilitates investigation of mRNA stability, processing, and translation controlled by ac4C.

Methodology:

ac4C-AFL determines optimal sequence length for model building, applies adaptive feature representation to extract RNA features, uses ensemble feature importance scoring with sequential forward search to select optimal features from 16 sequence-derived descriptors, constructs 176 baseline models with 11 classifiers, and applies two-step feature selection to integrate predictions from the most efficient models for final classifier training.