318

chain (Bayesian network; supervised machine learning). We learned about several applica­

tion examples in the book, such as for genome annotation, protein domain prediction, and

network regulation. For more details and information, see the article Sean R Eddy (2004;

What is a hidden Markov model? Nature Biotechnology volume 22, pages 1315–1316.

https://doi.org/10.1038/nbt1004-1315).

20.4

Modeling Metabolism and Finding New Antibiotics

Questions 4.1 to 4.5

One algorithm to calculate metabolic fluxes is elementary mode analysis. It calculates

enzyme chains that keep all internal metabolites in balance. That is, the enzymes consume

as much of an internal metabolite as other enzymes involved in that metabolic pathway

produce. External metabolites are source (e.g. glucose) and sink metabolites (e.g. pyruvate

as the end product of glycolysis), these cannot and do not need to be kept in balance.

Before starting the calculation, one makes a list of all enzymes and reactions, the stoichio­

metric matrix, which compiles the number of molecules each reaction consumes or pro­

duces. To assemble the metabolic enzymes and reactions correctly, one performs the

metabolic reconstruction. One looks over which enzymes should be present in the genome

based on the sequence analysis or completes this with further sequence analysis. Then you

can compile a list of all reactions and enzymes that are known for the metabolic pathway

(or metabolic network) you want to reconstruct in that organism. If I am careless and over­

look enzymes that are encoded in the genome, it may happen that individual reactions are

not connected to the metabolic network at all or that I assume wrong reactions that cannot

happen in the genome at all (best to always use and compare several databases). Enzymes

and reactions can be obtained e.g. from the KEGG database (https://www.genome.jp/

kegg/pathway.html; with EC numbers for all enzymes) and the ExPASy Biochemical

Pathways database (https://web.expasy.org/pathways). Enzymes found only in bacteria but

not in humans are potentially interesting antibiotic targets. Example software for meta­

bolic modeling is Metatool and YANAsquare/YANAvergence (faculty-owned software).

However, there are also other programs, e.g. CellNetAnalyzer (https://www2.mpi-­

magdeburg.mpg.de/projects/cna/cna.html).

Examples 4.6 and 4.7

A detailed description including a tutorial can be found at https://www.bioinfo.biozen­

trum.uni-­wuerzburg.de/computing/metatool_4_5/ or https://pinguin.biologie.uni-­jena.de/

bioinformatik/networks/metatool/metatool.html.

20  Solutions to the Exercises