P-TarPmiR accurately predicts plant-specific miRNA targets

Area of Science:

• Plant molecular biology
• Bioinformatics
• Genomics

Background:

• MicroRNAs (miRNAs) regulate gene expression post-transcriptionally, primarily studied in animals.
• Existing miRNA target prediction tools, often trained on animal data, perform poorly on plants.
• Plant-specific predictors struggle with non-canonical miRNA-mRNA interactions.

Purpose of the Study:

• To develop an accurate plant-specific miRNA target prediction method.
• To evaluate the impact of training data origin (animal vs. plant) on prediction accuracy.
• To provide a freely accessible tool for plant miRNA target prediction.

Main Methods:

• Retraining the TarPmiR predictor using the plant-specific TarDB database.
• Experimentally validating predictor performance across four plant species.
• Comparing predictors trained on animal data, plant data, and a combination.
• Developing the P-TarPmiR method excluding animal training data.
• Applying P-TarPmiR to predict targets in the soybean genome.

Main Results:

• Animal-trained miRNA target predictors show reduced performance on plant species.
• Accuracy of plant predictors improves with increased plant-specific training data.
• Excluding animal training data resulted in the most accurate plant-specific predictor.
• The P-TarPmiR tool demonstrates superior performance for plant miRNA target prediction.

Conclusions:

• Plant-specific training data is essential for accurate miRNA target prediction in plants.
• Animal-derived training data may hinder plant miRNA target prediction accuracy.
• The P-TarPmiR tool offers a highly accurate, plant-focused solution for miRNA target identification.
• Predicted soybean miRNA targets and GO enrichment analysis are publicly shared.