iPro-GAN

iPro-GAN employs generative adversarial learning to identify gene promoters and predict their transcriptional strength (strong versus weak), which influence RNA polymerase-mediated transcription initiation.

Key Features:

• Generative Adversarial Network: Uses a deep convolutional generative adversarial network architecture to learn complex sequence patterns.
• Two-layered classification: Implements a first layer for promoter identification and a second layer to classify identified promoters as strong or weak.
• Feature extraction: Applies the Moran-based spatial auto-cross correlation method to extract features from genomic sequences.

Scientific Applications:

• Disease Research: Enables analysis of promoter variations that may affect transcriptional regulation relevant to disease mechanisms.
• Gene Regulation Studies: Facilitates distinction between strong and weak promoters to study gene expression patterns and regulatory networks.

Methodology:

Benchmark and independent datasets were collected for training and testing; model training employed 10-fold cross-validation; reported accuracies on benchmark datasets were 93.15% for promoter identification (first layer) and 92.30% for promoter strength classification (second layer), and on independent datasets were 86.77% for identification and 91.66% for strength classification.