Notations and key principles of HomeoSplitter. Given a mapping on the contig (fragment) "ACCTGCT" one can count the nucleotides observed at each site for each accession (A). Questionable sites are those for which an excess of heterozygotes is observed (red arrows of A). Restricting the nucleotide counts to those questionable sites leads to the array Nb represented in (B). Even though homeologous copies are highly differentially expressed in each accession, considering them all at once here blur the signal. Indeed, at the second questionable site almost the same number of A (43) and G (44) are observed (C). To handle this problem, HomeoSplitter uses a specific expression bias for each accession; for instance considering the split defined by C
= [2,1] (i.e., pattern "CA") the estimated proportion of C
will be ~1/4 for the first accession (average of 5/20 and 10/42) and ~4/5 for the second one.