Ensemble-Learning and Feature Selection Techniques for Enhanced Antisense Oligonucleotide Efficacy Prediction in Exon...

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Antisense oligonucleotide (ASO)-mediated exon skipping is crucial for gene function studies and gene therapy.
• Current machine-learning predictors like eSkip-Finder are computationally intensive and have suboptimal accuracy.
• There is a need for more efficient and accurate computational tools for ASO efficacy prediction.

Purpose of the Study:

• To develop a computationally efficient and accurate method for predicting ASO efficacy in exon skipping.
• To improve upon existing machine-learning approaches by incorporating feature selection and ensemble learning.
• To reduce the computational burden associated with predicting ASO performance.

Main Methods:

• Implemented feature selection within machine-learning algorithms.
• Utilized ensemble-learning techniques, specifically a three-way-voting approach with random forest, gradient boosting, and XGBoost.
• Evaluated the approach on a dataset of experimentally validated exon-skipping events, including 2'-O-methyl nucleotides (2OMe) and phosphorodiamidate morpholino oligomers (PMOs).

Main Results:

• Achieved prediction computation times under ten seconds.
• Demonstrated improved prediction performance, indicated by higher R² values for both 2OMe and PMOs.
• Feature importance rankings from the model aligned well with previously published findings.

Conclusions:

• The proposed approach enhances the accuracy and efficiency of predicting ASO efficacy for exon skipping.
• This method has the potential to accelerate the development of novel therapeutic strategies using ASOs.
• Findings contribute to optimizing ASO design and advancing gene therapy approaches.