Del Bogotazo a la VII Conferencia de las FARC (1948 a 1982)

The procedure described in Figure 3.4 results in a draft metabolic network. Unfortunately, the series of mostly (semi-)automatic steps produces a reconstruction that is neither com- plete nor accurate (Iliopoulos et al., 2001; Devos and Valencia, 2001, 2000). Manual cura- tion, including a comparison of the network with knowledge from literature, is necessary to correct the errors and to fill in gaps. Some typical sources of errors are described below: • Incorrect or insufficient entries in the primary annotation- Genome annotations are subject to the limitations of the protein function prediction methods that were used in their creation. For example, OE2647F (VNG1148G in the NRC-1 strain) was originally annotated as an “amino acid kinase” (“ornithine carbamoyltransferase”). However, the fact that it is a conserved neighbor of mevalonate kinase (EC 2.7.1.36) suggested that it is somehow involved in the production of isoprenoids. Indeed, the ortholog in Methanocaldococcus jannaschii was found to phosphorylate isopentenyl phosphate.

• Newly characterized reactions (pathways)- It generally takes some time for databases such as KEGG to include newly characterized reactions (pathways). This is especially true if the reactions are limited to only a few organisms. For example, the classi- cal biosynthetic pathway for isopentenyl pyrophosphate (IPP) through mevalonate proceeds as follows: hydroxymethyl-glutaryl-CoA (HMG-CoA) is produced from acetoacetyl-CoA and acetyl-CoA (EC 2.3.3.10). HMG-CoA is reduced to mevalonate

(EC 1.1.1.34), which is phosphorylated to mevalonate phosphate (EC 2.7.1.36), which is converted to mevalonate diphosphate after a second phosphorylation step (EC 2.7.4.2), which is then finally decarboxylated to IPP (EC 4.1.1.33) (Figure 3.7). How- ever, in Methanocaldococcus jannaschii, it was demonstrated that the second phos- phorylation step is actually preceded by the decarboxylation step (Grochowski et al., 2006), rather than the other way around. Genetic evidence suggests that Halobac- terium salinarum likely uses this modified pathway (Falbet al., 2008), which at the time of writing is still not in KEGG. Accordingly, the “noncanonical” reactions had to be manually defined. Another example is the alternative pathway for aromatic amino acid synthesis described in Section 3.7.3.

• Inconsistencies and ambiguities between protein and function identifiers in different databases- Even if two databases already share at least one naming convention, prob- lems relating to the nature of the shared convention will still have to be dealt with. In the case of the EC system, several pathway repositories assign partially qualified EC numbers (partial EC numbers) to some reactions. For example, KEGG assigns the partial id of EC 3.1.3.- to the reaction (KEGG id R07280) in which 5-amino- 6-(5’-phosphoribitylamino)uracil is hydrolyzed and a phosphate group is removed. Accordingly, one may erroneously assign to this reaction any gene that satisfies the first three components of the EC number, as has been observed in several databases including KEGG (Green and Karp, 2005). The problem is that EC 3.1.3.-, by defi- nition, does not denote a specific reaction. Rather, it only specifies that the relevant enzyme is from a particular group, which in this case is the set of phosphoric mo- noester hydrolases. Situations like this need to be handled on a case by case basis. With respect to R07280, the two reactions before and after it in the linear segment of the pathway (riboflavin biosynthesis) in which it is involved have adequate genetic evidence in Halobacterium salinarum. Accordingly, it is likely that the reaction is also present in the organism. Certainly, the situation only gets worse if enzyme-gene synonym lists have to be used because the pertinent databases do not share naming conventions.

• Organism-specific cofactor requirements - Cofactor requirements are often organism- specific and have to be found elsewhere. The BRENDA database is a very convenient resource for this task.

• Unspecified reaction reversibilities - Reversibility information is not explicitly in- cluded in the KEGG databases. At best, KEGG uses single or bidirectional arrows in

its reference pathway images to indicate that a reaction is reversible or “irreversible”, respectively. These visual indicators can sometimes be inconsistent between path- ways (images). For these reasons, in this work we had to obtain reversibility in- formation from other sources. In particular, we used organism-specific data from literature as the primary basis. If this was not available, literature data from closely related organisms was used. In the absence of both, we used the assignments made in other reconstructions as well as the annotation by Ma and Zeng (Ma and Zeng, 2003), where the reversibility of each KEGG-defined reaction was assigned based on biochemical principles.

Figure 3.7: 2D Modified mevalonate path- way. One reason why manual curation is crucial to the reconstruction process is that it generally takes some time for databases to include newly characterized reactions (pathways). This is espe- cially true if the reactions are limited to only a few organisms. For example, isopentenyl pyrophos- phate (IPP) is the precursor of numerous impor- tant cellular constituents, including lipids and reti- nal. The classical biosynthetic pathway for IPP through mevalonate begins with the production of hydroxymethyl-glutaryl-CoA (HMG-CoA) from acetoacetyl-CoA and acetyl-CoA (EC 2.3.3.10). HMG-CoA is reduced to mevalonate (EC 1.1.1.34), which is phosphorylated to mevalonate phosphate (EC 2.7.1.36), which is converted to mevalonate diphosphate after a second phosphorylation step (EC 2.7.4.2), which is then finally decarboxylated to IPP (EC 4.1.1.33). However, in Methanocal- dococcus jannaschiiit has been demonstrated that the second phosphorylation step is preceded by the decarboxylation step (Grochowskiet al., 2006). Genetic evidence suggests thatHalobacterium sali- narum likely uses this modified pathway, which is currently not in KEGG. The red arrows cor- respond to the reactions of the classical pathway which are not in the modified pathway. The modi- fied steps are indicated in green. Putative enzyme- genes from Halobacterium salinarum R1 for each step of the pathway are provided in blue.