Computational Systems Biochemistry Group

Realistic predictions of flux distributions using constraint-based optimization requires reliable data that can be used to constrain the otherwise infinite solution space. Equilibrium constants or Gibbs free energy values may constrain the reversibility of a reaction, i.e. may determine whether the flux through its forward or backward direction is thermodynamically realizable. We recently published a method to directly include these data for the calculation of thermodynamically feasible flux distributions. As a main advantage, this method does not require a priori assumptions on the irreversibility of single reactions (Hoppe et al, 2007). However, thermodynamic data are not comprehensively available on genome-scale. Here, we propose a novel computational method to infer the unknown ΔG₀ value of a reaction from known ΔG₀ values of chemically similar reactions.

To quantify the chemical similarity of biochemical reactions, we have established a detailed classification procedure that assigns 3304 different chemical attributes to atomic groups occurring in presently characterized biochemical metabolites. Changes in these attributes between the substrate and product molecules are tracked on a per-atom basis and similarities between these reaction-specific attribute changes are assessed by the Tanimoto coefficient (T) assuming values between 0 (complete dissimilarity of reactions compared) and 1 (identity of reactions compared). Testing our method across a set of 1546 biochemical reactions, 216 of which being covered by experimentally determined ΔG₀ values - the root-mean-square distance (RMSD) between predicted and measured ΔG₀ values amounted to 8.0 kJ/mol, if a minimum similarity of T > 0.6 to reactions with known ΔG₀ values is assumed.

This value is significantly smaller than the RMSD of 10.5 kJ/mol achieved with the commonly used group contribution method (Mavrovouniotis, 1991; Jankowski et al., 2008). However, for less similar reactions, the group contribution method produces a more accurate prediction and a combination of both approaches is proposed. Clustering all reactions of a given metabolic network according to chemical similarity allows to identify minimal sets of reactions for which ΔG₀ values yet have to be experimentally determined in order to make reliable predictions of ΔG₀ values for the remaining reactions (manuscript in preparation).

Researchers

Sabrina Hoffmann
Dr. Kristian Rother
Dr. Andreas Hoppe
Sascha Bulik

Own Publications

Rother K, Hoffmann S, Bulik S, Hoppe A, Gasteiger J, Holzhütter HG. (2010) IGERS: Inferring Gibbs energy changes of biochemical reactions from reaction similarities. Biophys J., 98(11):2478-86. [PubMed]

References

Hoppe A, Hoffmann S, Holzhütter HG. (2007) Including metabolite concentrations into flux-balance analysis: Thermodynamic realizability as a constraint on flux distributions in metabolic networks BMC Syst. Biol.,1(1): 23. [pdf, journal, PubMed]

Jankowski MD, Henry CS, Broadbelt LJ, Hatzimanikatis V. (2008) Group contribution method for thermodynamic analysis of complex metabolic networks. Biophys J., 95(3):1487-99. [PubMed]

Mavrovouniotis ML. (1991) Estimation of standard Gibbs energy changes of biotransformations. J Biol Chem., 266(22):14440-5. [PubMed]

		Computational Estimation of Gibbs Free Energy Values of Enzymatic Reactions
Home Research VirtualLiver People Publications Software Positions Events Contact		Realistic predictions of flux distributions using constraint-based optimization requires reliable data that can be used to constrain the otherwise infinite solution space. Equilibrium constants or Gibbs free energy values may constrain the reversibility of a reaction, i.e. may determine whether the flux through its forward or backward direction is thermodynamically realizable. We recently published a method to directly include these data for the calculation of thermodynamically feasible flux distributions. As a main advantage, this method does not require a priori assumptions on the irreversibility of single reactions (Hoppe et al, 2007). However, thermodynamic data are not comprehensively available on genome-scale. Here, we propose a novel computational method to infer the unknown ΔG₀ value of a reaction from known ΔG₀ values of chemically similar reactions. To quantify the chemical similarity of biochemical reactions, we have established a detailed classification procedure that assigns 3304 different chemical attributes to atomic groups occurring in presently characterized biochemical metabolites. Changes in these attributes between the substrate and product molecules are tracked on a per-atom basis and similarities between these reaction-specific attribute changes are assessed by the Tanimoto coefficient (T) assuming values between 0 (complete dissimilarity of reactions compared) and 1 (identity of reactions compared). Testing our method across a set of 1546 biochemical reactions, 216 of which being covered by experimentally determined ΔG₀ values - the root-mean-square distance (RMSD) between predicted and measured ΔG₀ values amounted to 8.0 kJ/mol, if a minimum similarity of T > 0.6 to reactions with known ΔG₀ values is assumed. This value is significantly smaller than the RMSD of 10.5 kJ/mol achieved with the commonly used group contribution method (Mavrovouniotis, 1991; Jankowski et al., 2008). However, for less similar reactions, the group contribution method produces a more accurate prediction and a combination of both approaches is proposed. Clustering all reactions of a given metabolic network according to chemical similarity allows to identify minimal sets of reactions for which ΔG₀ values yet have to be experimentally determined in order to make reliable predictions of ΔG₀ values for the remaining reactions (manuscript in preparation). Researchers Sabrina Hoffmann Dr. Kristian Rother Dr. Andreas Hoppe Sascha Bulik Own Publications Rother K, Hoffmann S, Bulik S, Hoppe A, Gasteiger J, Holzhütter HG. (2010) IGERS: Inferring Gibbs energy changes of biochemical reactions from reaction similarities. Biophys J., 98(11):2478-86. [PubMed] References Hoppe A, Hoffmann S, Holzhütter HG. (2007) Including metabolite concentrations into flux-balance analysis: Thermodynamic realizability as a constraint on flux distributions in metabolic networks BMC Syst. Biol.,1(1): 23. [pdf, journal, PubMed] Jankowski MD, Henry CS, Broadbelt LJ, Hatzimanikatis V. (2008) Group contribution method for thermodynamic analysis of complex metabolic networks. Biophys J., 95(3):1487-99. [PubMed] Mavrovouniotis ML. (1991) Estimation of standard Gibbs energy changes of biotransformations. J Biol Chem., 266(22):14440-5. [PubMed]