[BioC] ChIPpeakAnno findVennCounts

[Ying Wu]

Dr Zhu,
After thinking about this for a couple days, I think the best way is a 
combined approach where more documentation is added to makeVennDiagram 
by such as adding findOverlaps to 'See Also' section and exposing the 
select option in findOverlappingPeaks as well as giving a 
warning/explanation when negative numbers appear. I think making the 
values 0 would simplify things too much for the end-user.

Personally, I think the following approach would be best for venn 
diagram but would require more calculations and the numbers would not 
add up.
100, 010, 001 - regions/peaks exclusive to one bed file (NOT total-overlap)
110, 101, 011, 111 - min(overlap) where overlap is calculated both bed1 
-> bed2 and bed2->bed1.

One case where I think 0s would be appropriate (but are negative numbers 
right now) is when you compare the same file but I have not quite 
figured out why this occurs yet.

Best,
Ying

On 3/23/2012 6:05 AM, Zhu, Lihua (Julie) wrote:
> Dai-Ying,
>
> Thanks for reporting this bug which happens when some of the peaks in a
> dataset overlap with multiple peaks in another dataset! What do you think if
> I set the number to 0 if it becomes negative? Thanks!
>
> Best regards,
>
> Julie
>
>
> On 3/22/12 9:49 PM, "Dai-Ying Wu"<daiyingw at usc.edu>  wrote:
>
>> Hi Dr Zhu,
>> Please disregard the accessing findVennCounts() section of the last email.
>> I have been able to reproduce the negative values in venn diagram with the
>> following code
>>
>> p1 = RangedData(IRanges(start=c(100, 500), end=c(199, 599), names=c("s1",
>> "s2")
>> p2 = RangedData(IRanges(start=c(110, 190, 320, 650), end=c(150, 199, 340,
>> 680), names=c("s1", "s0", "s2", "s3")))
>> p3 = RangedData(IRanges(start=c(130, 330, 530, 800), end=c(150, 350, 550,
>> 900), names=c("s1", "s2", "s3", "s4")))
>> vtest = makeVennDiagram(RangedDataList(p3, p2, p1), NameOfPeaks=c("p3", "p2",
>> "p1"), totalTest=100)
>>
>> if you flip the order around, it will not show negative numbers but the
>> venncount will be incorrect
>> vtest = makeVennDiagram(RangedDataList(p1, p2, p3), NameOfPeaks=c("p1", "p2",
>> "p3"), totalTest=100)
>> in this diagram, the overlap between p1 and p2 (110) should be 1 instead of 0
>> and (010) should be 1 instead of 2
>>
>> How do you advise I should proceed? Is there going to be a patch made to
>> makeVennDiagram to make the counting more consistent? I am currently working
>> on a dataset where I was looking at two replicates and pooled result and I was
>> getting negative numbers in the pooled result which led to very strangely
>> generated proportional venn diagrams.
>>
>> Best,
>> Ying
>>
>> ----- Original Message -----
>> From: Dai-Ying Wu<daiyingw at usc.edu>
>> Date: Thursday, March 22, 2012 5:30 pm
>> Subject: ChIPpeakAnno findVennCounts
>> To: julie.zhu at umassmed.edu
>>
>>> Hi Dr Zhu,
>>> I have been trying to troubleshoot a venn diagram that I tried to
>>> create with your very helpful ChIPpeakAnno software. I am trying to
>>> create a 3-way venn diagram from 3 BED files and the 3rd comparison
>>> gives a negative number. I have increased totaltest but that does
>>> not seem to be the issue. I then try and debug it by going into the
>>> code and try running it step by step, however, when i try to runb
>>> findVennCounts it gives an "Error: object 'findVennCounts' not
>>> found" and when I do "ChIPpeakAnno::findVennCounts" it gives not
>>> exported object from 'namespace:ChIPpeakAnno'
>>>
>>> If I do ?findVennCounts and try and use the example at the end of
>>> the page but it gives the same error. I am currently using
>>> ChIPpeakAnno 2.2.0
>>>
>>> Thanks,
>>> Ying Wu
>>>
>
>