Distinguishing the disease-associated SNPs based on composition frequency analysis.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Single-nucleotide polymorphisms (SNPs) are fundamental genomic variations.
• SNPs at microRNA binding sites can disrupt gene expression and are linked to diseases.
• Distinguishing disease-associated SNPs (dSNPs) from common SNPs is crucial for understanding disease mechanisms.

Purpose of the Study:

• To develop a computational method for accurately identifying disease-associated SNPs (dSNPs) related to microRNA modification.
• To differentiate dSNPs from common SNPs based on sequence characteristics.

Main Methods:

• Utilized sequence composition, transition, and distribution features, inspired by protein sequence analysis.
• Applied binary encoding for nucleic acid representation (A, T, C, G).
• Employed clustering for negative sample generation and random forest for model construction, optimizing feature subsets and sample ratios.

Main Results:

• Achieved prediction accuracy exceeding 90% on the testing dataset.
• Demonstrated practical applicability through promising external validation results.
• Principal component analysis confirmed the significant contribution of all engineered features to the model's performance.

Conclusions:

• The developed method effectively distinguishes disease-associated SNPs from common SNPs.
• The approach leverages sequence properties and machine learning for accurate dSNP identification.
• This work provides a valuable tool for investigating SNP-related diseases and microRNA regulation.