levels in gene ontology

Question

[ashish]

I am doing gene ontology using Blast2Go. Results look like this:

Level   GO ID   GO Name #Seqs
1   GO:0008150  biological_process  141
2   GO:0009987  cellular process    136
2   GO:0008152  metabolic process   68
3   GO:0044238  primary metabolic process   66
3   GO:0071704  organic substance metabolic process 66
3   GO:0044237  cellular metabolic process  62
4   GO:0019538  protein metabolic process   60
4   GO:0043170  macromolecule metabolic process 62
4   GO:0044260  cellular macromolecule metabolic process    58
5   GO:0036211  protein modification process    55
5   GO:0044267  cellular protein metabolic process  55
5   GO:0043412  macromolecule modification  55
6   GO:0006464  cellular protein modification process   55

Can someone tell me what levels should I use to make a informative bar graph. Here, things upto level 3 does not tell much about the genes I am studying.

thanks

[Benn]

Seems like you've answered your own question already...

[ashish]

Can you please elaborate a little more about levels. Is there any way to increase these levels to get better information.

[Petr Ponomarenko]

Why do you want to put different GO levels on the same bar graph, remove some data and provide no level/tree information? If you want to focus your information on a particular GO levels, you can make Y axis break from 70 to 130. This is a question of data representation, not data filtration. There is no need to loose data in order to make it comprehensible. Also, there is a Combined Graph option with Show parameter to make a graph representation for levels in Blast2GO. Or you can print the tree on the left of the bar graph as one figure like this

Answer 1

[Jean-Karim Heriche]

In general the higher the level, the more semantically specific the term is. However, the usefulness of levels is limited because different branches have different levels. For example, a very specific term could be level 4 in a branch but another term with similar specificity could be level 6 in another branch. This doesn't even take into account that some terms can belong to different branches at different levels.
If you want to make a graph summarizing your sequences, just pick terms covering the processes that you find relevant and collect the rest in an 'other processes' category.

[ashish]

This sounds reasonable. Thanks Jean.