Scikit-ribo Enables Accurate Estimation and Robust Modeling of Translation Dynamics at Codon Resolution

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• Ribosome profiling (Ribo-seq) measures protein translation but suffers from sampling errors and biases.
• Existing methods for estimating translation efficiency (TE) from Ribo-seq data are prone to inaccuracies, particularly for low-abundance genes.

Purpose of the Study:

• To develop an open-source package, Scikit-ribo, for accurate A-site prediction and TE estimation from Ribo-seq and RNA sequencing data.
• To address biases in TE estimation and improve genome-wide translation analysis.

Main Methods:

• Developed Scikit-ribo, an analysis package for Ribo-seq and RNA-seq data.
• Implemented a codon-level generalized linear model with ridge penalty for TE estimation.
• Validated TE estimates using mass spectrometry data for protein abundances.

Main Results:

• Scikit-ribo accurately predicts A-site locations and codon elongation rates (r = 0.99).
• The package corrects TE estimation biases for over 2,000 genes in S. cerevisiae, validated by protein abundance data (r = 0.81).
• Scikit-ribo enables direct determination of Kozak-like sequences from Ribo-seq data.

Conclusions:

• Scikit-ribo provides a robust framework for accurate genome-wide translation efficiency estimation.
• The package corrects known biases in Ribo-seq analysis and offers insights into translation regulation.
• Analysis of coverage requirements and cycloheximide treatment artifacts is provided for improved experimental design.