Full-length ribosome density prediction by a multi-input and multi-output model

Area of Science:

• Molecular Biology
• Computational Biology
• Genomics

Background:

• Translation elongation is a complex biological process regulated by intricate mechanisms in prokaryotes and eukaryotes.
• Ribosome profiling offers codon-resolution insights into translation but current computational models often overlook long-range codon interactions and the continuity of elongation.
• Modeling translation elongation at the full-length coding sequence (CDS) level remains an underexplored area.

Purpose of the Study:

• To develop a novel deep learning framework, RiboMIMO, for modeling ribosome density distributions across entire mRNA CDS regions.
• To capture correlations between neighboring and distant codons influencing translation efficiency.
• To identify key biological factors affecting translation elongation dynamics through interpretable analysis.

Main Methods:

• Developed a multi-input, multi-output deep learning model (RiboMIMO) to analyze full-length CDS ribosome profiling data.
• Incorporated analysis of correlations between codon translation efficiencies, considering both local and remote codon effects.
• Utilized an interpretable metric, the codon impact score, to assess individual codon contributions to elongation rates.

Main Results:

• RiboMIMO significantly outperforms existing methods in predicting ribosome density distributions along full-length mRNA CDS.
• The model successfully identifies known patterns of translation efficiency and reveals novel long-range codon associations.
• Demonstrated that codons distant from the ribosomal A site can influence translation elongation rate, a previously unreported finding.

Conclusions:

• RiboMIMO provides a powerful and accurate tool for studying translation elongation regulation at the whole CDS level.
• The discovery of long-range codon impacts offers new insights into the regulatory mechanisms governing protein synthesis.
• This work advances computational approaches for analyzing ribosome profiling data and understanding gene expression regulation.