Sequence-based prioritization of i-Motif candidates in the human genome.

INTRODUCTION: i-Motifs (iMs) are cytosine-rich, four-stranded DNA structures with emerging roles in gene regulation and genome stability. Despite their biological relevance, genome-wide prediction of iM-forming sequences remains limited by low specificity and high false-positive rates, leading to considerable experimental burden.

METHOD: To address this, we developed a refined computational approach that prioritizes high-confidence iM candidates using a Position-Specific Similarity Matrix (PSSM) derived from multiple sequence alignments. The human reference genome (hg38) was scanned using a custom regular expression targeting cytosine-rich motifs, followed by scoring each sequence with the PSSM. Statistical significance was assessed via permutation testing, one-sided t-tests, Benjamini-Hochberg correction, and Z-scores.

RESULTS: This pipeline identified 37,075 candidate sequences (15-46 nucleotides) with strong iM-forming potential. Validation against experimentally confirmed iMs and known G-quadruplexes (G4s) demonstrated significant differences in alignment scores and sequence similarity, confirming structural specificity. A random forest classifier trained on nucleotide features further supported the distinctiveness of the candidates, achieving a high classification performance.

CONCLUSION: This work presents a scalable and statistically robust method to enrich for biologically relevant iM sequences, providing a valuable resource for future experimental validation and the rational design of ligands targeting iMs to modulate gene expression in contexts such as cancer.