iDNA-ABF

iDNA-ABF predicts DNA methylation patterns from genomic sequences using a multi-scale deep biological language learning deep learning model to interpret sequential and functional determinants.

Key Features:

• Multi-Scale Deep Learning: Employs a multi-scale deep learning architecture that captures both sequential order and functional semantics from genomic sequences.
• Biological Language Learning: Uses biological language learning approaches to model genomic sequence patterns relevant to methylation.
• Interpretable Predictions: Integrates mechanisms that elucidate sequential determinants and provide interpretable insights into model predictions.
• Benchmarking Superiority: Demonstrates superior performance compared with existing state-of-the-art methylation prediction methods in comparative analyses.

Scientific Applications:

• Epigenetics: Predicts DNA methylation to support studies of epigenetic regulation and methylation-mediated control of gene expression.
• Genomics: Analyzes genomic sequences to identify methylation sites and their sequence-context determinants.
• Developmental Biology: Informs investigations of regulatory mechanisms shaping development via methylation patterns.
• Disease Research (including cancer): Supports analysis of methylation changes associated with disease progression such as cancer.
• Evolutionary Genetics: Aids comparative studies of methylation patterns across species or populations.

Methodology:

Trains a multi-scale deep biological language learning model on genomic sequences to predict methylation sites, processing large-scale biological data while capturing sequential order and functional context.