I am working with a dataset consisting of putative regulatory regions from mouse and human, part of which involves investigating the orthology relationships between elements identified in each species. I am interested in elements that are present in both genomes and those that are unique to one genome or the other. To identify these, I am using bnMapper, with the liftover chains from UCSC and reciprocal-best chains generated from them as described at http://genomewiki.ucsc.edu/index.php/HowTo:_Syntenic_Net_or_Reciprocal_Best. I use the reciprocal-best chains to identify orthologous elements and take the complement of the elements identied with the liftover chains to stringently identify elements that are unique to a species.
However, looking over the results, I am puzzled by the fact that a handful of sequences are mapping as orthologs using the reciprocal-best chains but not the liftover chains. As I understand it, the reciprocal-best chains should not include any alignment chains not present in the liftover chains they were produced from, and so this does not make sense to me.
My best guess is that this stems from gaps in the liftover chain being filled with lower-level chains during the netting process but I have not been able to verify this. I am wondering if anyone here can verify that this is the case and/or provide an alternative explanation for why this is happening. I have provided an example of an affected human region below. I am lifting between hg19 and mm9.
Thanks in advance!
chr17 40263777 40265157