GO DAG file format

The GO DAG file format is used by gomo_highlight and gomo. It is a quick parsing file, containing the structure and IDs from the Gene Ontology and can be automatically generated from the Gene Ontology's OBO format. A GO DAG file is packaged with the gomo databases available on the MEME Suite website.


File Structure

The file structure can be divided into 3 major parts.

  1. Header Comments
  2. Directed Acyclic Graph
  3. Graph Labels

Header Comments

The purpose of the header comments is to document the details of the OBO file that was used to generate it. Each line starts with a # symbol and the rest of the line can contain any content.

# Generated from an OBO file with the details:
# OBO Format Version: 1.2
# OBO Date: Fri Sep 25 11:40:00 EST 2009
# OBO Saved By: jane
# OBO Autogenerated By: OBO-Edit 2.1-beta1
# OBO Remark: cvs version: $Revision: 1.804 $

Directed Acyclic Graph

The purpose of the directed acyclic graph portion is to store the hierarchical structure of the Gene Ontology in a way that is quick to load into memory while still being compact in file size. The first line is the number of nodes in the DAG. Following that lines are in groups of 5 defining different attributes of the node. The group is on its own in the first line. The name is on the second line with the length coming first so memory can be preallocated for it. The node's position in the DAG is summarized in the third line with the total number of nodes above followed by the total number or nodes below. The fourth and fifth lines define the edges from the node to its parents and edges from the node to its children. Each line of edges starts with the edge count for that line which can be zero. Each edge is a number in the range zero to the node count minus 1 and is the index of the linked node. Values on the same line are tab separated. The order that the nodes are listed does not have any meaning other than providing a position to for the other nodes to link to.

node_count
node_1_group
node_1_name_length         tab node_1_name
node_1_nodes_above         tab node_1_nodes_below
node_1_parent_edge_count [ tab node_1_parent_1 ( tab node_1_parent_w )* ]
node_1_child_edge_count  [ tab node_1_child_1  ( tab node_1_child_x  )* ]
...
node_n_group
node_n_name_length         tab node_n_name
node_n_nodes_above         tab node_n_nodes_below
node_n_parent_edge_count [ tab node_n_parent_1 ( tab node_n_parent_y )* ]
node_n_child_edge_count  [ tab node_n_child_1  ( tab node_n_child_z  )* ]

Graph Labels

The purpose of the graph labels is to allow lookup of a graph node. Each line in the graph labels section has a symbol indicating if the label is the primary (>) or alternate (+) label, followed by a tab, followed by the label, followed by a tab, followed by the position of the associated node (index plus 1) or zero if the label is obsolete. The labels are ordered alphabetically.

label_count
> tab label_AA          tab position_of_node_AA
> tab label_AB          tab position_of_node_AB
> tab label_AC_obsolete tab 0
...
+ tab label_ZZ_alt_id   tab position_of_node_AB


Simple Example

Suppose there were the nodes A, B, C, D with the alternate names G,E,H,F and an obsolete name Z all from the grouping 'example' which I will shorten to x.

Node NameAlternate NamesParent NodesChild NodesGroupNodes AboveNodes Below
AGB, Dx03
BEACx11
CHBx20
DFAx10

So one possible output for this example would be as follows. Note that this example simulates tabs as tabs don't display properly in html.

# Simple example header
4
x
35   The full descriptive name of node C
2    0
1    3
0
x
46   Another descriptive name, this time for node A
0    3
0
2    3    2
x
38   Meaningful descriptive name for node D
1    0
1    1
0
x
21   Node B's name in full
1    1
1    1
1    0
9
>    A    2
>    B    4
>    C    1
>    D    3
+    E    4
+    F    3
+    G    2
+    H    1
>    Z    0


Generating from OBO files

If for some reason you can't source the GO DAG file from the MEME Suite website, then the tool obo2dag is provided in the scripts directory for the purpose of creating GO DAG files. It is an executable jar file with source packaged in the jar. As the OBO file format is still under active development we made use of a parser included with the OBO-Edit program which means our program is dependent on libraries from OBO-Edit. As this program is not likely to be needed by an end user we have not sought permission to include OBO-Edit's parser and so you will have to source the libraries yourself.

Required Libraries for obo2dag

The tool obo2dag is dependent on the OBO-Edit classes:

These classes have their own dependencies and it was discovered through a process of trial and error that the libraries needed from the OBO-Edit distribution are:

The current distributions of OBO-Edit seem to only include automated installers (no tar version) and so I recommend downloading the RPM version and using a program like 7-zip to extract the files you need. I found that the jar files were in the rpm at the location "/./opt/OBO-Edit2/runtime/". Once you have the jar files simply run obo2dag in the same folder.

Running obo2dag

As stated previously, obo2dag is an executable jar file and so if your system is setup correctly you can run it like you would a program and it will bring up the graphical user interface. To run the GUI from the command line type:
java -jar obo2dag.jar

If you need to run obo2dag in non-GUI mode then you must specify the class gomo.HierarchyParser and pass it the path to the obo file and the path to the output GO DAG file. The command is:
java -cp obo2dag.jar gomo.DAGParser <GO OBO File> <Output File>