Creating a Venn diagram based on protein frequency

[#1]

Hello all, 

    I have run into a quandary, and I feel confident there is a simple solution. I have sequenced several hundred proteins for three samples (actually, it's more, but let's start easy): A4-1, A4-5, and A4-8 I want to create a Venn diagram to depict the proteins that were uniquely expressed by certain samples vs. those that were shared. I have now essentially stacked all of the protein names into one column. The problem is, if I try to transpose to where the protein names are in the columns, it then merges those with the same name, which is NOT what I want. If I give each column a unique label, though, I need to go through and manually merge them, which could take a long time. I would like an output as in the attached image, with 0s and 1s represent absence and presence, respectively, of the protein listed in the column. I feel strongly that there is a way this can be done in under one minute using a feature of JMP Pro ver. 14 that I have never used. I tried "Text to columns," and I am thinking this might be one avenue, but I welcome any and all thoughts! Thanks, Anderson

[#2 · Accepted solution]

@abmayfield ,

I do not understand your data, especially how time and treatment should be factored into the analysis. I think Tables> Summary using a group and a subgroup will create the table of 0 and 1's you are looking for.  My interpretation of what you might be looking for is to use :Protein and :Sample as either the Group and Subgroup, respectively, or vice versa.

Below is a screenshot of using Summary with :Protein as Group and :Same as Subgroup. The JSL follows the picture.

Note that rows where N Rows is greater than 1 represent proteins found in more than 1 sample. and rows where N Rows equals 1 is a proteinn unique to that sample. The column Pattern is a contatenation of the character 0 and 1's and is a numeric representation of "areas" of a venn diagram. Running another table summary by Pattern would reveal how many are shared.

Data Table( "Ofav protein profiling stacked" ) <<
Summary(
	Group( :protein ),
	N,
	Subgroup( :sample ),
	Freq( "None" ),
	Weight( "None" )
)

  Summary by Pattern

Also, Tabulate might be useful to provide the lists of common proteins for each pattern, not just N.  This is the result of Tabulate, with Pattern and protein as grouping categories, only a small sample is captured below. This might not be exactly what you need, but hopefully provides some leads to your next steps.

[#3]

Brilliant! That is exactly what I wanted to do (despite not being able to properly elaborate it). I also forget about the tabulate function. The only thing I don't know how to do is create the "pattern" column, which will be very useful for making the Venn diagram. Did you just concatenate the prior columns or is there a specific "pattern" feature somewhere in the JMP package? Thanks so much for your help!

[#4]

@abmayfield ,

When wrote the other post, I used a column function (see below), however.... 

You can do this with Combine Columns.  Highlight the columns representing Sample values. Then from the Main Menu, select Cols, scroll down and select Utilities> Combine Columns and the the dialog below will appear, remove the default comma delimiter, uncheck Multiple Response and name the column Pattern, OK and JMP will create the column to the left of the first Sample column..

[#5]

Great! That is easy enough. In the end, my Venn diagram was way too complicated to be plotted (assuming proportional areas with 16 samples), but the table with the presence/absence data is the important thing I was after in my original post! Thanks for your help and hopefully this will be beneficial to others converting long lists of genes, proteins, or what have you, into numeric data.