BioPrediction-RPI: Democratizing the prediction of interaction between non-coding RNA and protein with end-to-end...

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Machine Learning (ML) algorithms are crucial for extracting knowledge from biological sequences in fields like healthcare and agriculture.
• Biological sequences are often categorical and unstructured, necessitating feature engineering for ML application.
• Current end-to-end ML pipelines require specialized user expertise, limiting broader adoption.

Purpose of the Study:

• To introduce BioPrediction-RPI, an end-to-end ML framework for identifying implicit interactions between biological sequences.
• To enable accurate sequence interaction prediction without requiring specialized ML expertise.
• To provide an interpretable report for users to understand prediction insights.

Main Methods:

• BioPrediction-RPI employs feature engineering to represent sequences using structural and topological features.
• Features are grouped to train partial models, with decisions combined for a final prediction.
• The framework includes an interpretability report for user insights.

Main Results:

• BioPrediction-RPI demonstrated competitive performance against expert-created models across 12 datasets.
• The framework achieved equal or superior performance in 40% to 100% of experimental cases.
• It showed capability for model fine-tuning with new data.

Conclusions:

• BioPrediction-RPI democratizes end-to-end ML by lowering the expertise barrier for biological sciences.
• The framework performs comparably to ML experts, enhancing accessibility and application.
• It facilitates the identification of complex sequence interactions, such as RNA-protein pairs.