Hi,
I edited the post to add additional informations
I did an experiment where I detect genomic events (mobile element integration - MEI) in different samples (10) divided in 3 groups across several time points (8). When a mobile element is integrated in a cell, it's irreversible. And each cell harbors only one integration. The probability to have the same integration point in two different samples is very (very) low. The position of integration is like a marker that can be traced in multiple time points. If the marker (i.e. the position of integration) is found in multiple time points, it show that this cell (or daughter-cells - as the integrated mobile element will be transmitted to all "clone" cells) is "recurrent" i.e. is being positively selected.
The data can be expressed as follow :
Sample 1
TP1 TP2 TP3 TP4 ...
MEI 1 1 0 1 1
MEI 2 0 0 1 0
MEI 3 0 1 1 1
Sample 2
TP1 TP2 TP3 TP4 ...
MEI 1 0 0 0 0
MEI 2 1 1 1 0
MEI 3 0 0 0 1
etc for the others samples...
0 = no event detected
1 = event detected
each event (MEI) represents a genomic position (chr-position). Most of the events are restricted to one samples. These data can be merged as follow :
#occurences
MEI 1 4
MEI 2 3
MEI 3 3
by counting the number of times we detect each event across all time points (if an event arises in the same TP of two different samples we counted it as one).
Now my problem is to perform a test that can compare the 3 groups in term of event recurrence (more an event occurs at different time point, more it's recurrent). A big problem in my experiment is that I could not detect all events due to the sensibility of the technic used. For example for the sample 1 :
Sample 1
TP1 TP2 TP3 TP4 ...
MEI 1 1 0 1 1
we should also see MEI 1 in TP2, but due to the limitation of the method we cannot...
Also if
TP1 TP2 TP3 TP4 ...
MEI 2 0 0 1 0
It's complicated to tell if event 2 appears in TP3, or before ...
In summary : How to test for event recurrence across different time points in samples from different groups
Thank you
There are a couple of things that you should clarify in your question:
- You seem to say that if an event is recorded at T1, it should be recorded at T2. This implies that the events are non-reversible, i.e. they don't change over time and the counts can never be decreasing. Is this correct ?
- What do you mean by "test for event recurrence across time points" ? The "more an event occurs at different time point, more it's recurrent" means to me that you want the sample(s) with the highest total. Or are you interested in how the events distribute over time e.g. time between occurrences ?
- Are you interested in processing each event separately or in all events simultaneously ?
Hi, 1) It's reversible, if an event is recorded at T1 it can disappear at T2. But in the other hand if I found an event at T1 and T3, it should also be present at T2. ok ? 2) With recurrent I mean an event that is found in multiple TP, thus "lives" longer than events with few or one TP 3) As events are mainly unique to each sample (it's very rare to find the same event in to different samples), it's better to take all events simultaneously (otherwise it will be complicated to compared the different samples..)
If it's reversible, how can you be confident that T1=1 and T3=1 implies T2=1 ? Can't you have a change between T1 and T2 and one between T2 and T3 ? If you're worried about false negatives, what about false positives e.g. in sequence (0,0,1,0) couldn't the 1 be an error ? If you want to go into error correction, you'll most likely need to develop a model of the process and its detection. I am still unclear on what you're trying to do. I understand that you want to compare 3 groups of samples but what is the question you're asking about these groups, what are the differences/similarities you're interested in ? You'll also need to clarify the context, for example are the events independent and does the current state of an event depend on any previous one ? It would probably be much easier if you told us exactly what the data is.
I edited the main question to try to answer your comments.