LncReader: identification of dual functional long noncoding RNAs using a multi-head self-attention mechanism

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Long noncoding RNAs (lncRNAs) can possess both protein-coding and noncoding functions, termed dual functional lncRNAs.
• These molecules play critical roles in fundamental cellular processes, making their identification crucial.
• Current experimental methods for identifying dual functional lncRNAs are laborious and cell-type specific, necessitating novel computational approaches.

Purpose of the Study:

• To develop an efficient in silico method for predicting dual functional lncRNAs.
• To introduce LncReader, a deep-learning model designed for accurate identification of these molecules.
• To provide a practical computational tool for accelerating dual functional lncRNA research.

Main Methods:

• Development of a deep-learning model, LncReader, utilizing a multi-head self-attention mechanism.
• Validation using benchmark datasets from the cncRNAdb project.
• Independent testing on in-house datasets generated through mass spectrometry proteomics, RNA-seq, and Ribo-seq in leukaemia cell lines.

Main Results:

• LncReader demonstrated superior performance compared to classical machine learning methods on benchmark datasets.
• Independent validation confirmed LncReader's accuracy and robustness in identifying dual functional lncRNAs.
• The deep-learning model achieved the best performance among tested computational tools.

Conclusions:

• LncReader offers an accurate and efficient computational solution for identifying dual functional lncRNAs.
• The tool overcomes limitations of traditional experimental methods, enabling faster discovery.
• LncReader facilitates advancements in understanding the roles of dual functional lncRNAs in cellular functions.