Transposons

Transposons identifies and analyzes 3'-end stem-loops in LINE-1 (L1) retrotransposons, Alu elements, processed pseudogenes, and mRNAs using machine-learning models that integrate sequence-based and structure-based features.

Key Features:

• Sequence-Based Models: Capture nucleotide compositional biases at 3'-ends of L1s, Alus, processed pseudogenes, and mRNAs to detect patterns indicative of stem-loop formation.
• Structure-Based Models: Incorporate physical, chemical, and geometrical properties of dinucleotides forming stems and position-specific nucleotide content in loops and bulges.
• Biophysical Parameters: Evaluate parameters including shift, tilt, rise, and hydrophilicity to characterize structural constraints of 3'-end stem-loops.
• Machine-Learning Integration: Combine sequence- and structure-derived features within machine-learning frameworks to improve recognition of 3'-end stem-loops.
• Detection Performance: Recognizes 62–68% of processed pseudogenes and mRNAs that contain 3'-end stem-loops.
• Cross-Element Detection: Applies to RNA/DNA secondary structures across multiple genomic elements beyond L1 and Alu.

Scientific Applications:

• Retrotransposition Studies: Supports investigation of retrotransposition mechanisms for LINE-SINE elements by detecting 3'-end stem-loops implicated in transposition.
• Genomic Element Analysis: Provides structural characterization of processed pseudogenes and mRNAs to reveal shared 3'-end stem-loop features despite sequence divergence.
• Potential mRNA Recognition: Identifies mRNAs with structural characteristics that could enable recognition by the L1 machinery.

Methodology:

Constructs machine-learning models trained on sequence and structural data, incorporating dinucleotide physical, chemical, and geometrical properties and position-specific nucleotide content; evaluated parameters include shift, tilt, rise, and hydrophilicity.