ILGBMSH: an interpretable classification model for the shRNA target prediction with ensemble learning algorithm.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Short hairpin RNA (shRNA)-mediated gene silencing is crucial for RNA interference (RNAi).
• Designing potent and reliable shRNA molecules is essential but challenging.
• Current biological methods for shRNA target selection are costly and time-consuming.

Purpose of the Study:

• To develop a precise and efficient computational method for designing potent and reliable shRNA molecules.
• To introduce an interpretable classification model, ILGBMSH, for shRNA target prediction.
• To provide biological insights into shRNA design through interpretable machine learning.

Main Methods:

• Utilized the Light Gradient Boosting Machine algorithm for shRNA target prediction.
• Extracted 554 biological and deep learning features beyond traditional sequence analysis.
• Employed Shapley Additive Explanations (SHAP) for model interpretability.

Main Results:

• ILGBMSH achieved state-of-the-art performance compared to existing shRNA prediction models.
• Feature analysis provided significant biological insights into shRNA efficacy.
• Independent experimental validation confirmed the model's predictive ability and robustness (Pearson's r = 0.985 for gene knockdown).

Conclusions:

• The ILGBMSH model offers a powerful and interpretable approach for shRNA design.
• This computational method significantly improves the efficiency and accuracy of RNA interference applications.
• The model's ability to provide biological insights aids in understanding the mechanisms of shRNA function.