Hi,

Thank you for providing this useful tool. I have used xTea to identify TE insertions in a large cohort and have a few questions regarding quality control and population-level merging.

I am using your x_vcf_merger.py script to combine nearby insertions together at the population level because, without merging, I end up with a very large number of rare insertions. I am only merging TEs with the same family (ex: only merge Alus with Alus, not Alus with LINEs).
Upon looking into what got merged together I noticed it combines TEs regardless of the reported SVLEN, which can vary greatly. For example, LINE1s should be around 6kb, but some of the LINEs that get merged into a single insertion can have a SVLEN from ~100bp to ~6kb.

I have also noticed that the x_vcf_merger.py script reports the most common SVLEN (correct me if I'm wrong), which is often shorter than the expected SVLEN for the TE family.

I had a few questions regarding this:

1. Would you recommend applying additional filtering based on SVLEN before running the merge script?
2. Should filters be applied based on subclass ("two_sides_tprt_both","one_half_side",etc) before merging? If so, where might I find the definition of each of these? From what I understand "two_sides_tprt_both" is the most reliable, but these aren't always the most common.
3. When I report SVLEN in my downstream analysis, should I use the SVLEN chosen by the x_vcf_merger.py script (even though it can be very small at times, even zero)? Or should I use the maximum SVLEN of the ones merged (which seems to be closer to the expected length based on TE family).

Thank you for your help and for developing xTea.