Cm-siRPred: Predicting chemically modified siRNA efficiency based on multi-view learning strategy.

Area of Science:

• Biotechnology
• Computational Chemistry
• Drug Discovery

Background:

• Rational modification of small interfering RNA (siRNA) is essential for developing effective siRNA therapeutics.
• Machine learning (ML) can optimize the design of chemically modified siRNA (cm-siRNA) by predicting efficiency, reducing development time and cost.
• Existing in-silico methods for cm-siRNA efficiency prediction face challenges with limited datasets, poor data representation, and lack of interpretability.

Purpose of the Study:

• To develop a robust and interpretable ML algorithm for predicting the efficiency of chemically modified siRNA.
• To address the limitations of existing in-silico methods in data representation and interpretability.
• To provide a practical tool for accelerating siRNA drug development through improved chemical modification design.

Main Methods:

• Developed the Cm-siRPred algorithm utilizing a multi-view learning strategy.
• Represented cm-siRNA using double-strand sequences, chemical modifications, and physicochemical properties.
• Integrated a cross-attention model for global correlation and a two-layer CNN for local feature learning.

Main Results:

• Cm-siRPred demonstrated exceptional performance in cross-validation and on an independent dataset.
• The algorithm showed robustness and generalization ability through case studies on approved siRNA drugs.
• A user-friendly webserver was developed for efficient cm-siRNA efficiency prediction and design assistance.

Conclusions:

• Cm-siRPred is a practical tool offering valuable technical support for siRNA chemical modification and drug efficiency research.
• The algorithm effectively assists in the development of novel small nucleic acid drugs.
• Cm-siRPred is freely available, promoting advancements in siRNA therapeutics research and development.