FastFeatGen

FastFeatGen accelerates parallel feature extraction from genome sequences and enables efficient prediction of DNA N6-methyladenine sites using machine learning methods.

Key Features:

• Parallel Feature Extraction: Implements shared memory parallelism to achieve approximately 10× speedup over sequential approaches for feature extraction from genome sequences.
• Exploratory Feature Selection: Employs a feature selection technique to identify more relevant features for input to machine learning methods.
• Extra-Tree Classifier: Uses Extra-Tree Classifier (ETC) as the machine learning method for prediction of N6-methyladenine sites.
• Multi-format Input Support: Supports text file input as shown in dataset.txt format, with fasta file compatibility via simple preprocessing.
• Query Prediction Service: Provides built models for direct prediction of query sequences without requiring model training.

Scientific Applications:

• DNA N6-Methyladenine Site Prediction: Predicts genomic locations of N6-methyladenine, an epigenetic modification involved in prokaryotic defense systems and human diseases.
• Rice Genome Analysis: Achieves 85.57% prediction accuracy on rice genome N6-methyladenine sites.
• Mouse Genome Analysis: Achieves 96.64% prediction accuracy on mouse genome N6-methyladenine sites.
• Large-Scale Genome Feature Extraction: Provides scalable parallel feature extraction for bioinformatics tools requiring processing of large genome sequence datasets.

Methodology:

FastFeatGen implements shared memory parallelism for feature extraction from genome sequences, followed by exploratory feature selection to identify relevant features, which are then processed by an Extra-Tree Classifier (ETC) for prediction of DNA N6-methyladenine sites.