KGB: A KEGG Browser KGB: A KEGG Browser Step by Step Tutorial

KGB: A KEGG Browser
Tutorial


The KEGG Browser (KGB) is developed for the purpose of extracting most information of interest from KEGG and generating reactions, i.e., chemical equations can be imported into the YANA (Yet Another Network Analysis) software package or any other desired tool which can read the raw (option 1) or tab-delimited text file (option 2). Subsequently users could perform analyses of the metabolism using the file.

KGB is completely implemented in Java, to obtain platform-independency. Using the KEGG API (Kyoto encyclopedia of genes and genomes application interface), KGB is capable to fetch the information on metabolism of a specific strain. The acquired information can be interpreted into a readable format by other functions of the KEGG API. We name this method as "online mode" in the program, since it relies mostly on remote processes. An alternative faster method is also implemented in KGB, named as "accelerated mode". This downloads the reaction database from KEGG in advance using a FTP protocol. The database is used as a dictionary to interpret the KEGG equation numbers locally. The latter method is highly recommended for studying a larger system, in order to enhance the work flow efficiency.

This tutorial describes a very typical working flowchart, which is supposed to fulfil most scientific purposes. Visitors could follow these steps to obtain a quick impression of this software.

I. Typical steps

1: Specify the organism which is going to be studied in the upper combo box (Figure 1a or Figure 1b).
2: Choose one pathway in the left table followed by a click on the "Reaction" button (Figure 2).
3: Select all the necessary equation items using radio buttons, verify their reaction directions, and dump them into the lower collection table (Figure 3). Repeat the step 2 and 3, until all the wanted equations have been ready prepared in the collection table.
4: Afterwards, users could use the "Dictionary" button to collect all the keywords of metabolites and enzymes, there a couple of buttons could aid users to shorten these keywords into a well-abbreviated form, the most useful two buttons are located in the middle panel (Figure 4).
5: Translate the reaction equations into abbreviated form and import them into YANA software package (Figure 5).

II. Example Figures


Figure 1a: Specify the organism. As the first step, users specify the organism of interest in the upper combo-box, here, lmo is the abbreviation of the organism Listeria monocytogenes.



Figure 1b: Specify the organism. Using the button , users invoke a browser to query KEGG about the organism terms, here regular expressions are supported by KGB, e.g., "e.coli", "e.co\w+" (any word starting with "e.co"). Once the organism of interest is selected, users should apply the "GO" button below the result list to pass to step2 "specify the pathway".


Figure 2: Specify the pathway. The pathway of interest should be chosen. Then users can apply the button to acquire all the involved equations (Step 2).


Figure 3: Specify desired enzyme reactions. In step 3, users could select all the necessary reaction equation items by checking the radio buttons, followed by verifying the directions of selected reactions. Metabolic maps (see Figure 7,8) can be used within KGB to accelerate this step. Repeat the step 2 and 3, until all the desired reactions are listed in the lower collection table. Users could save/restore them using the left buttons, the definition of these buttons can be found in the "additional features" chapter.


Figure 4: The upper Dictionary button calls a routine to generate a dictionary of abbreviated keywords. A very typical procedure is to apply two buttons (1 then 2 in the middle part), the program will collect the necessary metabolites and enzymes into the right dictionary table, where users are allowed to modify them manually, or process them automatically using the lower button panel.


Figure 5: Obtain a compact set of reactions. The final step 5 is translation. Applying the upper Translate button, the program will simplify all the collected equations according to the abbreviation dictionary (Fig. 4), which has been prepared in the step 4 before.


III. Additional Features

In this chapter, we describe some useful additional features and a couple of sophisticated tools involved in the software, they provide great flexibility (modelling of big and small networks in different organisms) and remarkably reduce searching complexity.

Figure 6: The editor buttons frequently appeared in the software.

Common buttons of KGB are shown in figure 6, from left to right they are responsible for changing the reaction directions, selecting all reactions, deselecting all, inserting a new row, erasing a selected row, eliminating one row, a data file restoration and saving the table content into disk using a tab delimited plain-text file.

The next three figures are snapshots of network illustration tools. They can be evoked by the buttons at the right side neighbouring the Translate "Translate" button. The first one is called "Figure" subroutine, which can immediately fetch the network figure from the KEGG server. Another choice is named "Map" subroutine, besides the network map provided by the first subroutine, it will mark in red the selected enzymes, so that the researchers can rapidly inspect their established system, e.g., whether all the necessary reactions have been covered or not. The last subroutine is designed to acquire plain-text information of enzymes and metabolites. It is named a "reference" browser, because this tabbed panel is in principle a mini web-browser, that allows people to click on all the EC numbers and metabolite names, so that the software will display the corresponding information.


Figure 7: Figure sub-tool. This visualizes the original KEGG predicted network, with green boxes of EC numbers denoting the enzymes that are present in the strain proteome.


Figure 8: The Map subroutine. This routine aids users to recognize the collected enzymes (in red) belonging to this pathway.


Figure 9: Reference browser. This provides reaction information according to the enzyme database.


IV. User tips and tricks

The KGB software provides two querying methods (named "Mode" in the menu), they apply different strategies when interpreting the KEGG information. The KEGG API enables us to obtain the latest information of strain-specific reactions, however, as the first step it will only give us rows of "encrypted-like" serial-numbers (KO number). Thus we have to translate them into a man-readable format. The default method is called " online mode" in the menu. It relies on the KEGG API to generate precise equations by remote calling. However, the process can be quite time-consuming for large sets of reactions. When users have many pathways to study, we recommend to take the alternative mode, named "Accelerated online" (Figure 10). It will inform the program to fetch the reaction database in advance using FTP protocol, afterwards, the reaction information can be translated locally instead of the default remote strategy.


Figure 10: Modes to fetch reaction equation. The modes to establish biochemical reactions can be switched in the menu. There is a remote call and a local procedure

Moreover, the local reactions file (as a database applied in accelerated mode) can be updated using a menu item (Update/Reactions, see Figure 11).


Figure 11: Reaction file update. KGB provides a function to update the reactions stored in the local reaction database file.

When translating and exporting the reaction abbreviations, some non-essential metabolites are filtered out as default. We named this "Filter" in the menu (Figure 12). This measure helps to reduce the complexity of the stochiometric matrix for later use, e.g. in YANA or Metatool.


Figure 12: Filter out metabolites which can be considered external. The filter removes metabolites which occur in stable pools (e.g. water).

Most of these metabolites (e.g., water, O2, CO2, H+) are relatively stable as a pool, so that they could be omitted in the established system. Additional metabolites can also be set as pool metabolites (and external). These are then omitted during translation and exportation, users assign a "^" symbol in the dictionary table as its abbreviated name. Afterwards, the translator can distinguish and eliminate them in all the related reactions. In rare case, some equations might be harmed due to this operation, i.e., one side or both sides of a certain equation will be empty after omitting metabolites. Users should pay attention to this point and re-validate the file after filtering.


V. Network simplification - some rules

In order to re-establish a larger metabolism system rapidly, besides using the helpful options from the last chapter, users may apply rules to simplify networks and their handling. These have also been successfully tested in the past usage and development of this software.

1. Avoid spelling mistakes

Keywords should be sorted and examined in the dictionary table: All the information provided by KEGG is stored in a plain-text format. By sorting and comparing the generated equations (from the previous step) in the dictionary table (operation: clicking the title and sorting the keywords in alphabet sequence), users can identify spelling mistakes (two neighbouring keywords are similar to one mismatch). If this happened by mistake, it is not necessary to change the original terms, the trick is then to assign then both the same abbreviation in the dictionary table, so that the software will treat them as the same metabolite or enzyme.

Mark external metabolites: In order to distinguish external metabolites in plain-text from internal ones, all the external metabolites should be given names ending with "(ex)" for external or "(extracellular)" for extra-cellular. The program will filter out these metabolites automatically and give them an external attribute when exporting the collection.

2. Use one enzyme for one reaction

When multiple enzymes catalyze the same reaction, you should keep only one of them, to simplify large systems (e.g. over 150 enzymes are involved).

3. Minimize reversible reactions.

If one reaction is changed from irreversible to reversible this can theoretically double the number of modes which have to be calculated. When one direction of the reaction is not supposed to be studied, or, e.g. not considered to happen as the following and previous reactions have been both recognized to be irreversible then give it only one direction.

4. Futile cycles should be minimized.

Like the previous rule, futile cycles can strongly boost the complexity, leading to high numbers of elementary modes to calculate. As the KEGG database yields abundant information on enzymes, some of them assemble to futile cycles. Some cycles are essential for the system, e.g., the well known TCA cycle. Some cycles are non-essential, they are composed by unique metabolites. Some reaction chains may be combined even into different cycles. In large systems, to avoid combination explosions, try to simplify these non-essential cycles to keep the system simple, in particular, remove where possible futile cycles.

5. Ex/internal metabolite rule.

In large systems, try to keep the number of external metabolites small. For instance, consider to minimize the different carbon/nitrogen sources. Thus many metabolic studies are based on minimum/basic medium, and glucose or sucrose are frequently used as C-source. To calculate growth or modes under such conditions consider to eliminate transporters and pathway chains for other sugars in large systems. Otherwise, these redundant carbon sources may dramatically increase the number of modes after performing an EM analysis.

In addition to establish the system of modes as simple as possible, the switch of an internal metabolite into external may considerably reduce the number of elementary modes. The detailed algorithm description can be found in "Dandekar et. al., BioSystem 2003:255-270". In addition, people can apply the C-program named "EI_EXCHANGE" to systematically determine whether an internal metabolite set as external simplifies the number of modes obtained.


VI. Sample files

The following sample files come with this tutorial:
S. aureus N317 carbohydrate metabolism system (central: composed by Glycolysis, Pyruvate metabolism, TCA cycle, PPP pathway, Fatty acid metabolism, etc.).
(Option1) Filename: sau_CH_G_collection
(Option2) Filename: sau_CH_G_FattyAcid_collection

S. aureus N317 biosynthesis system (mainly composed by above carbohydrate pathways and amino acid biosynthesis pathways).
Filename: sau_comp_G_collection

Step Hints:
1. Input one sample file using Restoration  (Fig 3)  
2. Click Dictionary , followed by Buttons 1,2 (Fig 4). Here users should assign "ex" (external) to metabolites, e.g., NAD+, NADP+, FAD+, to have "NAD+(ex)", "NADH(ex)", "NADP+(ex)", "NADPH(ex)", "FADH2(ex)", "FAD+(ex)", etc., in particular for the sample file of central carbonhydrate metabolisms. However, "NADP+", "NADPH" should be set as "internal" when the input file consisting of amino acid biosynthesis pathways is processed, because NADPH can be changed into NADP+ in this case.
3. Click Translate  to view the final result (Fig 5)

Detail description:
The upper demonstration files have been processed using the tricks (Chapter 4) and the rules (Chapter 5), moreover, the chemical terms were further simplified manually, to fit standard abbreviations. Users could skip step 1, 2 and load the sample file directly in step 3 (Figure 3) using the Restoration "restoration" button. Following the instructions of step 4 and 5, users should click the button 1, then 2 (Figure 4) to prepare an abbreviation dictionary. Alternatively, users could apply the button "automatic spell transformation" instead of button 2 (magician) after clicking the button 1 (collection). The software will still stay in this procedure and allow users to modify the abbreviations further. Here, as a suggestion, users could assign the "^" symbol to remove rare or general metabolites and metabolites of secondary metabolisms, for instance, pyrophosphate, orthophosphate, ferredoxin, protein, specific donors and hydrogen_sulfide could be marked and removed, otherwise, they might require many additional meatbolic steps and lead only to side-reactions. After clicking the Translate "Translate" button, a list of shortened equations appears in the table, users can export them either directly into YANA or save them as a general METATOOL input file.

The first file is a small system, it is flexible that can be easily calculated by YANA. The second demonstration system is rather huge and quite time-consuming when processed using an EM algorithm. Especially when there is no high performance computer available, users should set currency metabolites as external in large systems (Rule 5). The trick is to change e.g. NADH, NAD+ into external, since these are predicted to be stable pool metabolites if oxidative phosphorylation is present. Applying this rule, the calculation procedure can be further accelerated by considering urea as external.

Finally:A helpful tutorial can be found also within the software using the Help button.



Updated 30-Oct-2006, Chunguang Liang