search

geneontology / panther-enrichment

One of the main uses of the GO is to perform enrichment analysis on gene sets. For example, given a set of genes that are up-regulated under certain conditions, an enrichment analysis will find which GO terms are over-represented (or under-represented) using annotations for that gene set.
1 stars 2 forks source link

Panther slim query #3

Open kltm opened 7 years ago

kltm commented 7 years ago 
Follow on from:

http://jira.geneontology.org/browse/GO-1452

I was  curious about what the slim was showing.

in addition to the issues reported in the older ticket

The slim numbers for pombe seem  strange.
Below are panther reported and actual:
GO:0009987  cellular process       panther(1890) actual( 3983)
GO:0065007  biological regulation   panther(229) actual(1161) 
GO:0071840  cellular component org or bio panther(611) actual(1755)
GO:0032502  developmental process panther(25) actual(157)
GO:0051179  localisation panther(536) actual(1087)
GO:0008152  metabolic process panther(2082) actual(3147)
GO:0000003  reproduction panther(21) actual(335)
GO:0050896  response to stimulus panther(271) actual(677)

Other comments:

Should we have a pie chart view since we discourage this?
The histogram would be better, but the link for this is broken?
How are the slim terms devised?
Valerie Wood added a comment - 31/Jan/17 5:29 PM

From previous ticket:
Pie chart is very misleading for display of annotation numbers in these cases.
The legend for this pie chart is incorrect. It can't be percent of genes against total, because this would not be neatly 100% (and the total number of genes is more than the number of genes with component annotations, from the label at the top). It must be %annotations against total?
Valerie Wood added a comment - 31/Jan/17 5:33 PM

I attached an image of what this default slim looks like for pombe (approximately).
Hopefully this indicates why the pie chart view doesn't work for display of Genes->annotations
Seth Carbon added a comment - 31/Jan/17 5:41 PM - Restricted to go-curators

I'm looking for a ticket reference, but there was a lot of discussion about this way back when. According to my email, this was a "done" discussion already in 2009/2010--pie charts are misleading. E.g.:

We recommend that you avoid using pie charts because terms can have multiple parents with multiple relationships, therefore pie charts can be problematic to display your results as you invariably end up with a pie chart that adds up to more than 100%, which is nonsensical. A bar chart is a better alternative (e.g. see PMID: 12368864 fig. 3).

(Which is a 2002 ref, btw.)
Pie charts were removed in AmiGO fairly early on, replaced with bar charts (with the slimmer functionality removed completely later on when we moved to owltools and galaxy).

Cheers,

-Seth
Valerie Wood added a comment - 31/Jan/17 5:46 PM

+1.

I wrote about it in a pitfalls paper
https://www.ncbi.nlm.nih.gov/pubmed/18475267
and tell undergraduates on our GO courses not to use pie charts...

I'm very picky about it reviewing papers too....
Tanya Berardini added a comment - 31/Jan/17 8:55 PM

I'd like to redirect this to the appropriate Github tracker. Which one? Or should it go to the Panther tracker instead of a GO one?
Valerie Wood added a comment - 01/Feb/17 6:31 AM

Assigned to Huaiyu to tell us where the appropriate place is.
v
kltm commented 7 years ago

Sorry @ValWood, I could not get your image to upload here.

ValWood commented 7 years ago

(converted to jpeg) slim

thomaspd commented 6 years ago

You are right. In most places on the PANTHER site we support both the pie and bar chart view, so we should be able switch to the bar chart view quite easily.

huaiyumi commented 5 years ago

@kltm Is this issue resolvedd? If so, please close it. If not, can someone summarize the remaining problem? Thanks.

kltm commented 5 years ago

@ValWood Would have to answer--I only ported this from the old JIRA.

ValWood commented 5 years ago

I think the strange numbers for slim are also to do with the absence of "regulates" relationship.

Do you still have a pie chart (pie chart is bad for slimming because these are not discrete categories).

huaiyumi commented 5 years ago

@ValWood I will address the regulates relationship issue off line with you. Please do not populate this issue on every ticket. It is an issue that is more complicated than you think and really needs thorough testing before any decision can be made. Also we have both pie chart and bar chart. Pie chart is for annotations, not for genes. It is still useful, at least according to our users.

ValWood commented 5 years ago

Also we have both pie chart and bar chart. Pie chart is for annotations, not for genes. It is still useful, at least according to our users.

This is only because the users do not understand exactly what they are looking at. Yes the pie chart is for annotations but these are not discrete categories (they overlap with each other). This makes a pie chart uninformative, and we should not reinforce the myth that it is meaningful.

cmungall commented 5 years ago

I agree, pie charts for GO annotations are problematic.