Among structural variants, inversions are of particular interest in ecology studies as they can introduce phenotypic diversity and have a lasting impact on population genetics by locally impairing recombination [1] . Pangenome graphs (PG) are a way to represent all scales of genomic diversity in a species. Several tools have been developed to construct PG from genome alignments and to detect variants from the graph topology. PG were shown to be particularly efficient for identifying and genotyping deletions and insertions in model organisms [2] . However they have not yet been thoroughly assessed on inversion polymorphism.

We propose here to evaluate the ability of PG tools to represent and detect inversions in the case of a complex of butterfly species (Coenonympha genus, 9% nt divergence) for which we already detected a dozen of large (> 100 kb) inversions. To compare the tools, we selected a chromosome with 2 large, 6 medium sized (> 1 kb), and 9 small (< 1 kb) inversions, and built PG with 4 state of the art tools. Minigraph [3] and minigraph-Cactus [4] failed to build an accurate PG for such a degree of sequence divergence, while PGGB [5] and Cactus [6] found at most two inversions. In order to understand how PG handle inversions, we simulated the 17 known inversions in several synthetic chromosomes with increasing levels of single nucleotide divergence. We found that in such simplified graphs, most large simulated inversions are well represented with most tools. However, for smaller inversions and when the sequence divergence is higher, there is a significant variability in how the inversions are represented between PG tools. We analyzed the various obtained topological motifs, leading to methodological avenues for improving the detection of inversions in PG.