Fig 1.
Model constructing workflow.
Fig 2.
Sequence logo of the region from -30 to +30 in (A) first genes and (B) genes in the operon.
Fig 3.
Correlation between first genes’ PA/mRNA and their translation rates calculated by the Transim model.
Fig 4.
The correlation between translation rates and (A) PA/mRNA with mRNA expression level from normalized microarray data and (B) PA.
Fig 5.
Protein abundance per mRNA for first genes and subsequent genes in the operon.
Fig 6.
Differences in the expression level of three gene groups with different types of uAUG motif.
Fig 7.
Effect of features on the coding sequence to the protein expression level per mRNA.
Fig 8.
The correlation between PA/mRNA and the folding energy in the +1 to +30 region of the mRNAs.
Fig 9.
Effect of A. Protein sequence length, B. Protein half-life based on N-terminal rules, and C. The instability index (II) of protein.
Fig 10.
The correlation between PA/mRNA and (A) initiation rates and (B) elongation rates.
Table 1.
Correlation between new translation rate calculated by different machine learning models and PA/mRNA or PA on A. Testing dataset and B. Total dataset.
Fig 11.
The improved correlation between “New translation rate” and (A) protein expression and (B) protein expression per mRNA.
Fig 12.
Performance of SVR when predicting (A) protein expression and (B) protein expression per mRNA while integrating every single feature.
Fig 13.
The best Spearman correlation when combining from 0 to 10 sequential features to the model to predict A. PA/mRNA and B. PA.