DNAgenie

DNAgenie predicts DNA-binding residues in protein sequences and distinguishes interactions with A-DNA, B-DNA, and single-stranded DNA to enable DNA-type specific characterization of protein–DNA interactions.

Key Features:

• DNA-Type Specificity: Differentiates residue interactions with A-DNA, B-DNA, and single-stranded DNA.
• Custom Machine Learning Architecture: Employs a bespoke machine learning framework that uses comprehensive physiochemical profiles derived from input protein sequences.
• Two-Step Refinement Process: Applies a two-step refinement mechanism to reduce cross-predictions where residues interacting with other ligands are misidentified as DNA-binding.
• Performance Superiority: Demonstrates improved residue-level prediction accuracy on independent test datasets, with the refinement step particularly effective at reducing cross-predictions.
• Protein-Level Predictions: Aggregates residue-level outputs to coarse-grained protein-level predictions that compare favorably with recent tools distinguishing double-stranded versus single-stranded DNA binders.
• Human Proteome Analysis: Empirical analysis on human proteome sequences shows substantial overlap with known DNA-binding proteins and identifies several hundred candidate putative DNA binders.

Scientific Applications:

• Protein function annotation: Annotates protein function from sequence by identifying DNA-binding residues and their DNA-type specificity.
• Discovery of novel DNA-binding proteins: Identifies candidate novel DNA-binding proteins in proteome-scale analyses.
• Mechanistic studies of protein–DNA interactions: Facilitates exploration of residue-level molecular mechanisms underlying interactions with A-DNA, B-DNA, and single-stranded DNA.
• Genomic regulation research: Supports studies of genomic regulation by mapping DNA-type specific binding properties of proteins.

Methodology:

DNAgenie extracts physiochemical profiles from protein sequences, applies a custom machine learning architecture, and uses a two-step refinement mechanism to generate DNA-type specific residue-level predictions.