Fig 1.
Complex interactions of various components in cell metabolism.
Multi-omics data has provided the quantitative readouts of these components, which helps us to elucidate the interactions among the multi-layer regulations.
Fig 2.
Four modules are designed to implement omFBA algorithm: 1) transcriptomic-phenotype data collection (Step 1~2), 2) “phenotype match” algorithm (Step 3~5), 3) omics-guided objective function in FBA (Step 6~7), and 4) phenotype data validation (Step 8).
Fig 3.
“Phenotype match” algorithm for low and high glucose conditions.
The simulated and observed ethanol yields matched well for low (A) and high (B) glucose conditions, respectively. Negative correlations between the weighting factors of minimizing the overall enzyme usage and the observed ethanol yield were found for low (C) and high (D) glucose conditions.
Fig 4.
Correlation between phenotype-matched weighting factors and gene expressions.
The absolute values of the correlation coefficients in one of the training datasets were ranked from high to low (only the top 30 genes were shown here). The top 3 genes were chosen as the genetic markers to derive the omics-guided objective function (blue bars).
Fig 5.
Prediction accuracy of omFBA algorithm.
Direct comparison of the predicted and observed ethanol yields in low (A) and high (B) glucose conditions. The omFBA algorithm was repeated for 40 times and the percentage of matched predictions of omFBA algorithm were calculated and ranked for low (C) and high (D) glucose conditions.
Fig 6.
Effect of cutoff p-value on omFBA prediction using “small pool” of genes.
Three cutoff p-values, i.e., 0.05, 0.67, and 0.95, were used to filter the transcriptomics data. For each cutoff p-value, we re-ran the omFBA algorithm for 40 times and calculated the percentage of matches between the predicted and the observed ethanol yields in the validation datasets.
Fig 7.
Four key flux ratios (PGI/G6PDH2, FBA/TKT1, ENO/PPCK, and PYK/PDC) were selected to be correlated with phenotype-matched weighting factors, observed ethanol yields, and the ratios of the corresponding gene expression levels for low and high glucose condition. All the values of the ratios were exponential. Abbreviations: PGI, glucose-6-phosphate isomerase; G6PDH2, glucose 6-phosphate dehydrogenase; FBA, fructose-bisphosphate aldolase; TKT1, transketolase; ENO, enolase; PPCK, phosphoenolpyruvate carboxykinase; PYK, pyruvate kinase; PYRDC, pyruvate decarboxylase.
Table 1.
Key flux ratios compared with previous studies using 13C metabolic flux analysis.