The L. major Friedlin genome is 32.8 Mb in size, with a karyotype of 36 chromosomes. The genome data (version 5.2) was downloaded from the Sanger Institute public database ftp://ftp.sanger.ac.uk/pub/databases/L.major_sequences/. The Pathway Tools component of PathoLogic was used to build the initial version of LeishCyc from the genomic data 25. PathoLogic requires the sequence for each genomic element (a chromosome in this case), and associated annotation file. The chromosome sequences were extracted from 36 corresponding Sanger XML files, and were edited and reformatted with in-house developed programs. The L. major genome annotation originally provided by Sanger was in the Artemis format, and we used Artemis to convert the annotation file for each chromosome into the GenBank format. The resulting GenBank files were edited to add headers and subsequently the automated trial and build procedure was performed in PathoLogic.

The L. major genome contains 8283 genes predicted to encode polypeptides. The Pathway Tools software matched 574 of these genes to MetaCyc reactions based on the presence of an EC number in the gene annotation. Another 267 polypeptides, where no EC number was supplied, were matched to reactions based on their annotated name. The automated build resulted in 841 enzymes predicted for L. major. After the initial build, the list of 'probable enzymes' was constructed. Probable enzymes were gene products predicted to be enzymes but which could not be matched to any particular reaction by PathoLogic. These entries were manually reviewed and assigned to reactions where possible. There were 328 probable enzymes predicted for L. major and 148 were assigned to reactions by manual review.

Extensive manual curation of the database was performed based on literature search, and in-house experimental and bioinformatics studies. This included verification of enzymes and reactions deduced from the original genome annotation, refinements and improvement in the annotation of genes, enzymes, reactions and pathways, assignment of evidence codes, and inclusion of literature citations. At present LeishCyc contains 1027 enzymes and 566 metabolites organized into 704 enzymatic reactions, 37 transport reactions, and 143 metabolic pathways (Table 1).

Only pathways present in MetaCyc can be automatically incorporated into the pathway database by Pathway Tools 25. As MetaCyc contains predominantly bacterial and plant pathways, some pathways known to be present in Leishmania spp. were not present in the initial LeishCyc build. For example, MetaCyc lacked pathways involved in the assembly of the major surface glycoconjugates of Leishmania, including the biosynthesis of glycosylphosphatidylinositols (GPIs) and related glycolipids, and the assembly of complex phosphoglycans on the cell surface and secreted proteins and glycolipids 26. These and other new pathways were therefore manually created for LeishCyc based on the literature references. In addition, it was necessary to modify some of the automatically imported pathways in order to accurately represent known metabolic pathways in Leishmania spp. For example, the pathways for dolichyl-diphosphooligosaccharide and fatty acid biosynthesis were modified to reflect what has been experimentally observed or predicted for Leishmania 2728. In total, 66 pathways were created or modified in LeishCyc [see Additional file 1]. We have also added links between the LeishCyc pathways to show how they connect to each other.

After the initial build of LeishCyc, it was necessary to review the pathways and remove false-positive predictions 19. All pathways were reviewed, and those deemed to be supported by weak evidence were removed. For example, a pathway was removed if it did not contain any enzymes that were unique to the pathway and there was no experimental evidence for the pathway existence in the Leishmania spp. In some cases, pathways were deleted and replaced with Leishmania-specific pathways. For example, two pathways for phospholipid biosynthesis were present after the initial build (phospholipid biosyntheses I and II). These were replaced with Leishmania-specific pathways for phospholipid biosynthesis (ester phospholipid biosynthesis and ether phospholipid biosynthesis) 2930. In total, 128 pathways were removed from LeishCyc after the initial build [see Additional file 2].

Our own experimental work was used to add and verify some of the information present in LeishCyc. For example, GC-MS analysis of polar metabolites from cultured L. major promastigotes revealed the presence of several metabolites (i.e. glucitol and glycerol 2-phosphate) for which exogenous sources or biosynthetic enzymes were lacking, indicating the presence of new or unanticipated reactions. Recent analyses of sugar phosphates using high resolution Fourier-transform ion cyclotron resonance (FT-ICR) mass spectrometry, identified a novel mannose cyclic phosphate that is the primer for the major intracellular reserve carbohydrate of Leishmania, linear polymers of mannose which we have now termed mannogen 3132. While none of the enzymes involved in the assembly of the mannogen primer or downstream steps have been identified, the biochemically delineated steps have been incorporated into LeishCyc.

Targeted bioinformatics studies were used to aid curation and improve the LeishCyc annotations. For example, the original genomic annotation implied only two enzymes to participate in the pathway 'dolichyl-diphosphooligosaccharide biosynthesis'. Literature review has shown that this pathway is indeed present in Leishmania spp. 27, and hence our bioinformatics studies were directed towards identifying genes coding for missing enzymes of this pathway. We used hidden Markov models (HMMs) to identify the L. major genes encoding each of the mannosyltransferases in this pathway (ALG1, ALG2, ALG3, ALG9, and ALG11). Sequences that had been characterised in other organisms were used to build HMMs for each gene product. These models were then used to scan predicted L. major proteins to identify the most likely candidate for each individual ALG gene. The functional assignments made based on bioinformatics studies were documented in the annotations with the appropriate evidence codes (see below).

The functions of a number of Leishmania spp. genes have been identified in the literature since the L. major genome was published. As a result, these genes were not accurately annotated in the LeishCyc automated build which relied on the original annotation of the L. major genome project. We used extensive searches and manual reviews of published literature to incorporate additional Leishmania genes, proteins, enzymes, and transporters in LeishCyc. If a gene had been identified in L. major, the published accession number was used to identify the gene in LeishCyc. In some instances, we have judged the quality of the published information and entered the information accordingly. For example, 25 new annotation refinements were proposed for the L. major genome based on weak similarity using BLAST searches 7. One of these genes (LmjF31.1780) was identified as 'sphingosine N-acyltransferase' based on the similarity to Cryptococcus neoformans sphingosine N-acyltransferase (E-value of 4 × 10^-6) although a protein BLAST search of the NCBI non-redundant database returns a list with over a hundred hits with similar or better E-value, including Trypanosoma brucei and Trypanosoma cruzi proteins of unknown function (E-value of ~4 × 10^-70). In such cases, where we believed that further evidence is needed to firmly support the proposed annotation, we have quoted the literature source and the proposed annotation, while retaining the original 'unknown' function associated with the database entry. In cases where it was deemed that computational evidence was sufficiently strong, the new functional annotation was introduced with the appropriate evidence code, as described below.

In addition to identifying published L. major genes, we also identified L. major orthologs of enzymes and transporters that have been characterized in other Leishmania species and trypanosomatid species such as T. brucei and T. cruzi. The L. major orthologs were identified by a systematic Needleman-Wunsch alignment of the given sequence against L. major predicted proteins with a processing pipeline built in-house. The results of Needleman-Wunsch alignment were manually reviewed for highest similarities and, if deemed appropriate, the L. major gene encoding the protein was annotated as the predicted ortholog (see below for the explanations of the evidence code assignments). In such cases, the percent sequence identity and/or similarity that the L. major sequence shared with the known homologous sequence was recorded in the LeishCyc annotation entry. For example, the myo-inositol transporter (MIT) has been characterized in L. donovani, but had not been identified L. major. The Needleman-Wunsch alignment of the L. donovani sequence against all L. major peptides identified LmjF24.0680 as the clear candidate for this protein in L. major, and the peptide with the greatest similarity to L. donovani MIT. This gene was originally annotated as 'sugar transporter, putative' in both the original GenBank annotation file and in the GeneDB entry for LmjF24.0680. We manually annotated LmjF24.0680 as the predicted L. major MIT, and assigned the evidence code indicating that the inference was computational. In addition, for every gene in LeishCyc, we have added a link to the corresponding entry in GeneDB which directly connects the entries from the two databases.

Literature searches have identified a number of genes that have been linked to a particular enzyme or transporter in Leishmania spp. or other trypanosomatids. Such genes were identified in LeishCyc, checked as to whether they were linked to the correct reaction(s), and, if not, the respective entries were corrected.

Some enzymes were associated with an incorrect EC number in the L. major Genome Project annotation file, resulting in the enzyme being linked to incorrect reaction(s). For example, the phosphomannomutase enzyme catalyzes the reaction EC 5.4.2.8 (α-D-mannose 1-phosphate → mannose 6-phosphate), but in the annotation file it was associated with EC 3.1.3.11 and thus was linked to the EC 3.1.3.11 reaction (fructose 1,6-bisphosphate + H₂O → fructose 6-phosphate + P_i). In the subsequent manual curation, phosphomannomutase was removed from the EC 3.1.3.11 reaction and linked to the EC 5.4.2.8 reaction. In total, we found 7 enzymes to be annotated with the incorrect EC number in the L. major genome file, and we linked these enzymes to the correct reactions.

For genes not associated with an EC number in the L. major genome annotation file, the Pathway Tools software has attempted to link the gene based on matches of the product name to enzyme function. For example, the gene LmjF15.1010 was annotated as glutamate dehydrogenase and was matched to three glutamate dehydrogenase reactions each with a different EC number (EC 1.4.1.2, EC 1.4.1.3 and EC 1.4.1.4). The three reactions only differ in the cofactor used:

EC 1.4.1.2: L-glutamate + NAD⁺ + H₂O → α-ketoglutarate + ammonia + NADH

EC 1.4.1.3: L-glutamate + NAD(P)⁺ + H₂O → α-ketoglutarate + ammonia + NAD(P)H

EC 1.4.1.4: L-glutamate + NADP⁺ + H₂O → α-ketoglutarate + ammonia + NADPH

LmjF15.1010 is the predicted ortholog of the L. tarentolae mitochondrial glutamate dehydrogenase 33. The L. tarentolae enzyme uses NAD⁺ as a cofactor, but also has an NADP⁺ binding site, and so, in this case, the linkage of LmjF15.1010 to EC 1.4.1.3 was kept, and removed from EC 1.4.1.2 and E.C. 1.4.1.4.

In certain cases, manual intervention was required to link enzymes to the multiple reactions they catalyze. For example, the enzyme pteridine reductase was associated with EC 1.5.1.33 and was thus automatically linked to only one reaction. However, this enzyme has been shown to catalyze additional reactions in folate and biopterin metabolism in L. major 34, and we manually linked pteridine reductase to these additional (three) reactions. Similarly, the enzyme trypanothione synthetase was associated with EC 6.3.1.9 and automatically linked to one reaction. However, it has been demonstrated experimentally that this enzyme also catalyzes EC 6.3.1.8 35, thus we linked trypanothione synthetase to this reaction as well.

For newly discovered enzymes without annotation in the original genome project files, the required enzyme objects were manually linked with the relevant reactions and, if necessary, the reaction in question was created. For example, the L. major inositol phosphorylceramide (IPC) synthase gene was recently identified by 36. This gene was listed as a hypothetical protein in the L. major genome annotation file. We changed its annotation in LeishCyc to 'inositol phosphorylceramide synthase', manually created a new reaction (ceramide + L-1-phosphatidylinositol → inositol phosphorylceramide + 1,2-diacylglycerol), and linked the product of the gene to this reaction.

In addition to metabolic enzymes, a concerted identification and curation of transport reactions was performed. After the automated build, there were 23 transporters identified in LeishCyc, but only 6 transport reactions. Furthermore, many of the identified transporters had not been assigned specific transporter identities (e.g. MIT). Using the literature, we identified a further 26 L. major transporters (making 49 in total) and created 31 transport reactions [see Additional file 3].

The BioCyc ontology allows evidence codes to be assigned to support assertions in the BioCyc type database 37. If the supporting evidence was experimental, an evidence icon of a flask appears in the Pathway Tools software visual representation, with the assigned evidence code being 'EV-EXP' (see Figure 1). In LeishCyc this evidence code was manually assigned to signify experimental evidence, when the evidence came from any of the Leishmania spp. If the supporting evidence was only computational the evidence code 'EV-COMP' was assigned (this type of evidence is shown as a computer icon). If the evidence is supported by a publication, alongside each evidence code there is a link to the supporting publication. The code for curated proteins shows the evidence that supports the association of the protein with its linked reaction (i.e. for enzymes, this is the evidence that the enzyme catalyzes a given reaction or, for transporters, that the protein transports a particular substrate). In cases where the L. major ortholog was identified by the LeishCyc curator from similarity to a published sequence, the evidence code assigned was 'EV-COMP-HINF-FN-FROM-SEQ' (human inference of function from sequence), with additional explanations and the percent identity that the L. major sequence shares with the published sequence given. Currently, 208 proteins (including 2 protein complexes), 254 reactions, and 130 pathways have been assigned evidence codes in LeishCyc, and 200 references have been added to the database (Table 1). An example of a LeishCyc metabolic pathway with evidence icons and codes as displayed by the Pathway Tools software is shown in Figure 1.

Figure 2 shows alterations in protein expression levels during the differentiation of L. donovani promastigotes to in vitro differentiated (axenic) amastigotes 40, mapped onto the LeishCyc pathways. Enzymes that are decreased in the amastigote stage are shown in yellow while those that are increased are shown in red. Decreases in expression levels of enzymes involved in glycolysis (3 enzymes) and the pentose phosphate pathway (6 enzymes) are apparent, as are increases in enzymes involved in gluconeogenesis (2 enzymes), oxidative phosphorylation (4 enzymes), amino acid catabolism (3 enzymes), and fatty acid β-oxidation (5 enzymes). Additional patterns in the data become apparent in this representation, including decreases in some enzymes involved in nucleotide biosynthesis (8 enzymes), and increases in enzymes of ergosterol biosynthesis (9 enzymes). The use of LeishCyc greatly reduced the time needed to produce representations of these stage-specific metabolic changes. Furthermore, the Omics Viewer can be used to display the time points as a progressive series of images, creating an effect of data animation [see Additional file 4].

In the second example, LeishCyc was used to visualize data from metabolic profiling experiments performed in our group (Figure 3). L. mexicana MZ 379 promastigotes were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated foetal bovine serum (iFBS) at 27°C. Log phase promastigotes were harvested approximately two days after inoculation of the media. Axenic amastigotes were generated from stationary phase parasites (5–6 days after passage) by adjusting the conditioned media to pH 5.5 with HCl and the addition of iFBS to 20% 4142. The adjusted culture was incubated at 33°C and amastigote-like forms of the parasite were harvested on days 5 and 6. Parasite metabolism was quenched and polar metabolites extracted, derivatized, and analyzed by gas chromatography-mass spectrometry (GC-MS), as described previously 43. Figure 3 shows a colour-coded representation of changes in metabolite steady-state concentrations in axenic amastigotes relative to log-phase promastigotes. Significant changes in the steady state levels of many metabolites were detected. In particular, the levels of hexose phosphates and intermediates in glycolysis were reduced, while levels of many amino acids increased. Such a representation of metabolomic data in the context of the cellular biochemical pathways is highly useful for the observation of patterns in relative changes. In addition, the built-in capabilities of Pathway Tools allow one to interactively interrogate the organism metabolic pathways map overlaid with experimental data (such as those shown in Figures 2 and 3), in order to investigate observed patterns.

Chokepoint analysis has been used to prioritize potential drug targets in the P. falciparum PlasmoCyc database 22. Chokepoints are defined as reactions that consume unique substrates or produce unique products, potentially important criteria for drug target prioritization 2244. Pathway Tools was used to identify 324 chokepoint reactions in LeishCyc. These reactions have 145 enzymes associated with them, corresponding to 132 genes [see Additional file 5]. This list includes a number of enzymes previously predicted to be essential for normal growth or infectivity. Included in the list is lanosterol 14α-demethylase (LmjF11.1100), the protein target of ketoconazole, the only clinical drug for leishmaniasis with an identified protein target 45. The TDR Targets database http://www.tdrtargets.org46 was used to identify 31 genes in this list that do not have human orthologs [see Additional file 5]. Interestingly, a number of the genes identified in the chokepoint analysis of LeishCyc were also identified in a corresponding analysis of P. falciparum 22.