DeepShape: estimating isoform-level ribosome abundance and distribution with Ribo-seq data

Area of Science:

• Computational biology
• Molecular biology
• Genomics

Background:

• Ribosome profiling (Ribo-seq) is crucial for understanding translation.
• Mapping Ribo-seq reads to transcripts and assigning multi-mapped reads to isoforms are key challenges.
• Current methods for handling multi-mapped reads are suboptimal, leading to potential inaccuracies.

Purpose of the Study:

• To develop an RNA-seq-free computational method for accurate ribosome profiling.
• To estimate isoform-specific ribosome abundance and ribosome density profiles.
• To analyze translational regulation in cancer cells.

Main Methods:

• Developed DeepShape, a deep learning model for Ribo-seq data analysis.
• DeepShape estimates ribosome abundance and profiles without requiring RNA-seq data.
• Introduced the Codon Residence Index (CRI) to analyze ribosome speed at the codon level.

Main Results:

• DeepShape demonstrated superior accuracy in simulations compared to existing methods.
• Applied DeepShape to Ribo-seq data from PC3 human prostate cancer cells.
• Identified distinct translational regulation patterns for different isoforms of invasion/metastasis genes.
• Observed PP242 treatment-specific regulation of codon translation.

Conclusions:

• DeepShape is a powerful and accurate tool for Ribo-seq data analysis.
• The method facilitates a deeper understanding of translational regulation.
• DeepShape advances the analysis of ribosome profiling data.