Guidelines for extracting biologically relevant context-specific metabolic models using gene expression data

Modelling dynamic host-pathway interactions at the genome scale

“…Our work contributes to the emerging area of hybrid metabolic modelling that attempts to expand the predictive power of traditional approaches and incorporate a broader range of cellular processes relevant for pathway engineering. Recent examples include the use of machine learning to augment the predictive power of GEMs 59,60 , the integration of FBA and bioprocess models 61 , and expanding the boundaries of GEMs to include signalling pathways 62 . We anticipate this area will continue to grow and draw inspiration from machine learning techniques as practitioners strive to balance mechanistic and data-driven approaches to model biological systems 63 .…”

Multi-omic characterization of antibody-producing CHO cell lines elucidates metabolic reprogramming and nutrient uptake bottlenecks

“…Context-specific models for the pool and clone cultures for each phase were extracted using the workflow described earlier (Gopalakrishnan et al, 2023). First, the measured uptake and secretion rates were imposed as bounds in the i CHO1766 genome-scale metabolic model (Hefzi et al, 2016) and the inactive reactions were identified using flux variability analysis (FVA) (Mahadevan and Schilling, 2003).…”

“…All metabolites with available extracellular measurements were included in the set of core reactions. Since mCADRE shows the least variability in model content (Gopalakrishnan et al, 2023), ensembles of 20 models were generated in each case and then combined to account for potential alternate solutions. Model similarity was quantified using the Jaccard Index defined as follows: …”

“…First, we analyze the overall nutrient consumption and product yield in the bioreactor and show that high-producing clones do not necessarily need to grow to high cell densities, and that low cell density clones can achieve high titers through high specific antibody productivity. We built cell line-specific metabolic models with gene expression data using a published extraction workflow (Gopalakrishnan et al, 2023) using the mCADRE algorithm (Wang et al, 2012) and StanDep (Joshi et al, 2020) for expression thresholding and observed differences in model content involving lipid metabolism, glycosylation, and intracellular transport pathways. Since energy- and precursor-producing pathways (e.g., central metabolism, amino acid catabolism, and nucleotide biosynthesis) were conserved across all cell lines in all phases, we probed intracellular flux distributions using Monte-Carlo flux sampling after overlaying the uptake and secretion rates inferred from exometabolomics data on to the constructed phase- and cell line-specific models and found that glycolysis was primarily driven by glucose uptake and the TCA cycle was driven by glutaminolysis and degradation of acetogenic amino acids such as lysine and valine.…”

Genome‐scale metabolic models applied for human health and biopharmaceutical engineering

“…On the other hand, pruning methods identify a set of candidate reactions and then remove them one by one while maximizing the number of reactions removed without impacting the model's ability to simulate known metabolic processes. There have been comprehensive comparisons and evaluations of these different methods [55][56][57]. Generated specific models from these methods have been widely applied in simulating the metabolic phenotypes under particular genetic or environmental perturbations for various applications related to a myriad of biomedical advancements [58].…”