EnhancerNet: AI for predicting gene enhancer functionality

This study introduces EnhancerNet, a model for classifying normal and tumorous gene enhancers in colorectal cancer. It compares one-hot encoding with Position-aware Oligonucleotide Composition Descriptor-Nucleotide Density (POCD-ND) encoding. One-hot encoding achieved higher accuracy and specificity, while POCD-ND excelled in recall, suggesting its potential for detecting tumor-associated enhancers. Our results suggest that encoding performance may vary depending on the dataset characteristics and the architectural design of the model, highlighting the importance of dataset-specific optimizations in enhancer prediction tasks.