How to draw Venn digram for two groups peaks

[lee_sh]

Hi, everyone. I meet a problem in drawing Venn diagram to describe the overlapping of peaks between two groups. I defined one base overlapping between one peak from A and another peak from B as overlapping. Now, the problem comes. Sometimes, one peak in A group may overlapping with two near peaks in B group. Then , the number in A group overlapped with B group is not as same as that in B group. How to solve this? Thank you very much!

[westerman]

I guess I do not understand the problem. There will be X number of elements in set A that have overlaps with set B. There will be Y number of elements in set B that overlap with set A. X and Y do not need to be the same number in order to create a Venn diagram.

Now if you want a more complex relationship; e.g., something along the lines of element 'I' in set A has a relationship to elements 'J' and 'K' in set B ... well then a Venn diagram is not what you are looking for.

[DZhang]

Lee_sh,

My suggestion is to exclude the "ambiguous" peaks from your lists first, then check 1) the Venn Diagram without the ambiguous peaks; 2) the number of ambiguous peaks. If it represents a small percentage, you may ignore. If it is substantial, you need to fine tune your separation and comparison algorithm.

[kmcarr]

I guess I do not understand the problem. There will be X number of elements in set A that have overlaps with set B. There will be Y number of elements in set B that overlap with set A. X and Y do not need to be the same number in order to create a Venn diagram.

Rick,

I'd have to respectfully disagree with that statement. The overlapping portions of a Venn diagram contain the elements which sets A and B have in common. Since it is single set of elements who size is a finite integer N; set A contains all N elements and set B contains all N elements.

Lee_sh's problem is that a Venn diagram is not an appropriate representation for the data as defined. It would be appropriate to display peaks which are common to the two sets, with the stipulation that a peak in set A can only be considered in common with a single peak of set B and vice versa.

[lee_sh]

I guess I do not understand the problem. There will be X number of elements in set A that have overlaps with set B. There will be Y number of elements in set B that overlap with set A. X and Y do not need to be the same number in order to create a Venn diagram.

Now if you want a more complex relationship; e.g., something along the lines of element 'I' in set A has a relationship to elements 'J' and 'K' in set B ... well then a Venn diagram is not what you are looking for.

Good answer. Thanks. Maybe I can also use the percentages of each set to draw Venn diagram.

[lee_sh]

Lee_sh,

My suggestion is to exclude the "ambiguous" peaks from your lists first, then check 1) the Venn Diagram without the ambiguous peaks; 2) the number of ambiguous peaks. If it represents a small percentage, you may ignore. If it is substantial, you need to fine tune your separation and comparison algorithm.

Thanks. The "ambigous" peaks(like one peak in A overlaps two peaks in B) are accepted as one possiblity in my study. Someone suggested me to use gene instead peak for comparison.

[lee_sh]

Rick,

I'd have to respectfully disagree with that statement. The overlapping portions of a Venn diagram contain the elements which sets A and B have in common. Since it is single set of elements who size is a finite integer N; set A contains all N elements and set B contains all N elements.

Lee_sh's problem is that a Venn diagram is not an appropriate representation for the data as defined. It would be appropriate to display peaks which are common to the two sets, with the stipulation that a peak in set A can only be considered in common with a single peak of set B and vice versa.

Yes. Strictly, Venn diagram may be not suitable for this question if the elements in two sets can not meet the criteria as you discribed. However, the phenomena of one peak overlapping with more is common in ChIP-seq especially histone modifications. Any better method to describe such relationship between two sets?

[westerman]

Rick,

I'd have to respectfully disagree with that statement. The overlapping portions of a Venn diagram contain the elements which sets A and B have in common. Since it is single set of elements who size is a finite integer N; set A contains all N elements and set B contains all N elements.

You are, of course, correct. I am not sure what I was thinking. I was probably over-thinking because, as you say below...

Lee_sh's problem is that a Venn diagram is not an appropriate representation for the data as defined...

[kmcarr]

Yes. Strictly, Venn diagram may be not suitable for this question if the elements in two sets can not meet the criteria as you discribed. However, the phenomena of one peak overlapping with more is common in ChIP-seq especially histone modifications.

I don't dispute that the data are typical for the type of experiment performed. My comments were strictly to do with the presentation of the data.

Any better method to describe such relationship between two sets?

What's wrong with a nicely constructed table?

[rory]

We tend to look at this in two ways. One is to look at the percentage of peaks in set A that overlap ones in B, and vice-versa (percentage in B that overlap A). The other way is to merge the peaks, which gives you a common set of (wider) peaks. In your example where two peaks in one set overlap single peak in the other set, you would be left with a single peak that covers all three of them. Then you can do a proper Venn diagram.

There are function in our Bioconductor package, DiffBind, that make it easy to do both of these: dba.plotVenn will do the merging and plotting in one step, and dba.overlap will give you the each-way overlap rates of unmerged peaks (and extract the overlapping and non-overlapping sets as well).