Questions and Answers

1. What is GO?
2. What is GO Terms Classifications Counter?
3. What is DAG?
4. What this tool is designed for?
5. How does it work for me?
6. How are the results presented to me?
7. What's the file format for upload?
8. Can I use my own classification method?
9. How are ancestral classifications counted?
10. What is Transitive Closure?
11. What are the "single occurences" and "all occurences"?
12. Why my sum of the counts is more than the total GO terms in the input?
13. What are "counted terms" and "odd terms"?
14. How to properly interpret my results? Any hints to make better use of the tool?
15. Is there a way to improve my counting results?
16. How may I cite the use of this GO Classification Counter tool?

1. What is GO?

   GO is short for Gene Ontology which uses a controlled vocabulary to describe gene and gene product attributes in any organism, and to describe the basic term categories and relationships.

2. What is GO Terms Classifications Counter?

   The GO Terms Classifications Counter is comprised of several perl CGI programs coupled with a MySQL DBMS that stores the GO terms DAG data (pre-computed terms associations).

3. What is DAG?

   DAG is short for Directed Acyclic Graph, a directed grap with no directed cycles, that is, for any vertex v, there is no nonempty directed path that starts and ends on v. DAGs can be considered to be a generalization of trees in which certain subtrees can be shared by different parts of the tree. In a tree with many identical subtrees, this can lead to a drastic decrease in space requirements to store the structure. The Figure on the right shows a simple DAG structure (a) and how the the distances between terms may be calculated (b).

4. What this tool is designed for?

   Originally this tool was created to meet the needs in large scale EST analysis where people wish to get a rough idea how their annotated ESTs are distributed in different gene categories. This tool also finds its utility in understanding microarray data where large number of up/down regulated genes may be quickly categorized to show the trends of changes in terms of which gene classes are changed most. This can be used in combination with, for example, pathway analysis.

5. How does it work for me?

   The program takes input of GO terms (e.g. GO:0007067) in a list format or unformatted plain text file, and allows user to choose one of the available classification methods such as GO_slim, GOA, EGAD, MGI_GO_slim, ... or GO-ROOT (You can upload your own "slim" classification to use. However, if you have a new "slim" classification that you don't want to upload every time, and wish to be made useful to others as well, let us know and we will be happy to add it), performs the count, and returns the results on the web (if it takes too long, it will email the user with the link to results). The output includes the counts and percentages in a sorted tabular form, plus a pie chart.

6. How are the results presented to me?

   Shown on the right is a snap shot of the main part of a CateGOrizer results page where the counted GO terms are shown in tabulated form along with a pie chart. The raw GO IDs are also appended below as a reference, separated in "counted" and "odd" groups. Each GO term and GO class in the graph is hyper- linked to AmiGO browser at the GO web site for details (Click on the graph to see a real example).

   ( Click on the graph for a larger picture )

7. What's the file format for upload?

   The GO Term files for upload must be in plain text (ascii) format, prefereably laid out as a list, one entry per line, as shown in the following example.

   GO:0008629
   GO:0051241
   GO:0007157
   GO:0016849
   GO:0004383
   ... ...
   

   However this format is not strictly enforced because the program is "smart enough" to fish things out if the file is not formated as a single column list.

8. Can I use my own classification method?

   Sure you can. Your classification file must be in a list, plain text (ascii) format, one line per entry, led with GO_ID and separated from its definitions with a tab or space, as in:

   GO:0009058	biosynthesis
   GO:0008152	metabolism
   GO:0009056	catabolism
   GO:0019725	cell homeostasis
   GO:0005515	protein binding
   GO:0005840	ribosome
   ... ...		... ...
   

9. How are ancestral classifications counted?

   The transitive closure are built in GO database schema and the paths are pre-computed from every node to all of its ancestors, which is equivalent to computing the path from every node to all of its descendents (See reference "ii"). The count of a term into to a parental term is simply the process to find if the path lead its way to the ancestor.

10. What is Transitive Closure?

    The relationships between GO terms are not in simple tree structure (although sometimes partial structure may appear to be like hierarchies), but are organized in structures called directed acyclic graphs (DAGs). For example a GO term may have more than one child, and/or have more than one parent (See reference "i"). The transitive property of numbers states that if A = B and B = C, then A = C. Derby applies this property to query predicates to add additional predicates to the query in order to give the optimizer more information. This process is called transitive closure.

11. What are the "single occurences count" and "all occurences count"?

    It is common that multiple paths exist between an ancestral/parental term and a child term. Single occurences count is that we count only once when multiple paths are found between an ancestral term and a child term. The Single occurences count is often used to get an idea how the child terms are "classified". It can often effectively avoid an inflated total counts. In contrast, when each and all paths between an ancestral term and a child term are counted, we call it all occurences count. This can be useful to get an idea how complex the terms are related.

12. Why my sum of the counts is more than the total GO terms in the input?

    One may ideally wish a Term belongs to, and is counted into, an ancestor term only once. However in reality this is often not the case. As discussed above, a GO term can trace back to more than one ancestor parents, and therefore the trace of a child term to an ancestor term may form more than one independent path, making a (child) term counted multiple times into an ancestor term. Consider situation where multiple consecutive relationship exists along a path, and such relationship "fork" might happen to any of the element terms, thus the overall sum of counts may appear much "inflated". If such "inflation" is significantly larger than acceptable, one may want to scrutinate the GO Terms in the classification used, see if their representative coverage overlap, and do some necessary manipulations of the counting results to improve the represenation (see answers to the next question).

13. What are "counted terms" and "odd terms"?

    Along with the counted results, the Counter lists the "Counted Terms" and "Odd Terms". The "Counted Terms" are those that are found belonging to at least one of the classes in a given classification method. The "Odd Terms" are not found belonging to any classes in the given classification method. Both groups of terms are listed in a sorted order, with the frequence of a term in the raw GO Term data set placed in parenthesis postfixing each term. This information maybe helpful for users to evaluate how the GO annotations are represented under the given classification method. For example, if there are many "odd terms" to a classification method, it may imply that the method may not be the best fit for the raw data set. Users are encouraged to try out different classification methods or modify the classification to improve the representation of the raw data set. The classification that produce the least number of "odd terms" may be the best for your data set.

14. How to properly interpret my results? Any hints to make better use of the tool?

    As the count of GO terms into each ancestral term is independent of one another, the counted results may be selectively used, and percentages recalculated as long as the selected terms represent a well covered spectrum of your scope. This helps to avoid redundancy, and may make best sense for a particular data set. For example, in the "GO_Slim" classification, the three main GO classes, "molecular_function", "cellular_component", and "biological_process", may themselves already constitute a full coverage of the GO spectrum, therefore overlap with other (child) terms in the "GO_Slim" in terms of representative coverage. Therefore the counts of these three terms may be taken out, and the sum of remaining terms can still be taken as "100%", and the percentages for each term recalculated.

15. Is there a way to improve my counting results?

    Yes. As shown in the following graph for an example, Comparison of two analysis of one raw GO id data set, showing the importance of properly selecting a classification method. In (a), a "immune class" was chosen as the classification method; while in (b) the "GO slim" was the classification method.

    ( Click on the graph for a larger picture )

    A closer examination at a classification counting results on 3,216 GO terms, showing the importance of carefully selecting a suitable "GO Slim" to use. In (a), three root GO classes: "biological_process", "molecular_function", and "cellular_component" were included in the classification method, which caused much inflation of the total counts due to the overlaps between parental classification terms. In (b), when these three classes were excluded from the classification method, the counts appear to be "more normal".

    (a)

    (b)

    GO Class ID	Definitions	Counts	Fractions		GO Class ID	Definitions	Counts	Fractions
    GO:0008150	biological_process	971	12.74%		GO:0008150	biological_process	971	39.17%
    GO:0003674	molecular_function	780	10.23%		GO:0003674	molecular_function	780	31.46%
    GO:0005575	cellular_component	728	9.55%		GO:0005575	cellular_component	728	29.37%
    GO:0005623	cell	589	7.73%		Total		2479	100.00%
    GO:0005488	binding	354	4.65%		GO Class ID	Definitions	Counts	Fractions
    GO:0005622	intracellular	290	3.81%		GO:0005623	cell	589	7.73%
    GO:0008152	metabolism	250	3.28%		GO:0005488	binding	354	4.65%
    GO:0003824	catalytic activity	207	2.72%		GO:0005622	intracellular	290	3.81%
    GO:0005515	protein binding	192	2.52%		GO:0008152	metabolism	250	3.28%
    GO:0007154	cell communication	179	2.35%		GO:0003824	catalytic activity	207	2.72%
    GO:0005737	cytoplasm	161	2.11%		GO:0005515	protein binding	192	2.52%
    GO:0005886	plasma membrane	143	1.88%		GO:0007154	cell communication	179	2.35%
    GO:0005576	extracellular region	137	1.80%		GO:0005737	cytoplasm	161	2.11%
    GO:0007165	signal transduction	135	1.77%		GO:0005886	plasma membrane	143	1.88%
    GO:0004871	signal transducer activity	110	1.44%		GO:0005576	extracellular region	137	1.80%
    GO:0007275	development	96	1.26%		GO:0007165	signal transduction	135	1.77%
    GO:0016043	cell organization and biogenesis	94	1.23%		GO:0004871	signal transducer activity	110	1.44%
    GO:0016787	hydrolase activity	84	1.10%		GO:0007275	development	96	1.26%
    GO:0019538	protein metabolism	79	1.04%		GO:0016043	cell organization and biogenesis	94	1.23%
    GO:0005615	extracellular space	75	0.98%		GO:0016787	hydrolase activity	84	1.10%
    ...	...	...	...		...	...	...	...
    Total		6621	100.00%		Total		4142	100.00%

16. How may I cite the use of this GO Classification Counter tool?

    A menuscript is in preparation. In the meantime you are welcome to cite the use of this tool as:
    Zhi-Liang Hu, Jie Bao and James M. Reecy (2008) CateGOrizer: A Web-Based Program to Batch Analyze Gene Ontology Classification Categories. Online Journal of Bioinformatics. 9 (2):108-112.

1. The Gene Ontology Consortium (2006). An Introduction to the Gene Ontology. URL: http://www.geneontology.org/GO.doc.shtml, Last modified: February 17, 2006.
2. Chris Mungall, (2004). GO Database Schema. URL: http://www.godatabase.org /dev/sql/doc/godb-sql-doc.html, Last modified: April 5, 2004.
3. Hu, Zhi-Liang, Jie Bao and James M. Reecy (2007). A Gene Ontology (GO) Terms Classifications Counter. Plant & Animal Genome XV Conference, San Diego, CA, January 13-17, 2007.
4. Hu, Zhi-Liang, Jie Bao and James M. Reecy (2008). CateGOrizer: A Web-Based Program to Batch Analyze Gene Ontology Classification Categories. Online Journal of Bioinformatics. 9 (2):108-112.

By Zhiliang Hu
Associate Scientist
Dept of Animal Science
Iowa State University