Identification of sequence motifs significantly associated with antisense activity

Area of Science:

• Molecular Biology
• Bioinformatics
• Genetics

Background:

• Predicting antisense oligonucleotide (ASO) suppression activity is crucial for rational nucleic acid design.
• Understanding sequence properties associated with ASO activity is essential for developing effective predictive models.
• Identifying non-significant properties can streamline model development and improve efficiency.

Purpose of the Study:

• To identify sequence motifs significantly associated with high or low antisense suppression activity.
• To analyze the properties of these significant motifs.
• To evaluate the performance of predictive models using these motifs as features.

Main Methods:

• Randomization procedure to discover significant sequence motifs.
• Analysis of motif properties, including thermodynamic characteristics.
• Support vector machine (SVM) modeling using significant motifs as features.

Main Results:

• Discovered 155 motifs associated with high and 202 with low antisense suppression activity.
• Motifs range from 2 to 5 bases and exhibit thermodynamic properties consistent with existing research.
• No correlation found between motif position and antisense activity; significant motifs often appeared as subwords of others.
• Support vector regression (SVR) models using significant motifs showed increased correlation compared to models using all possible motifs.

Conclusions:

• Thermodynamic properties of significant motifs reinforce the link between probe/target thermodynamics and antisense efficiency.
• The independence of motif position and activity simplifies modeling, enhancing efficiency and reducing overfitting.
• Increased SVR correlation with significant features suggests factors beyond thermodynamics influence antisense efficiency.