As the a priori selection of the topologies could be thorny, we decided to achieve AU-tests that relax this assumption. The site-wise lolikelihoods of each tree was estimated from the concatenated dataset using PAUP and applying a GTR+I+C model. Log-likelihoods of site-pattern trees were then used to calculate the p-values for the AU- and the SH- tests in CONSEL. Sequences obtained in this study were deposited in EMBL with the following accession numbers JN418213-JN418217. For mt DNA, 992 bp of the cytb gene were determined for two samples using 12 to 15 overlapping fragments. Only two mt fragments were amplified for CH476 and none were obtained for CH560. Because mt DNA is present in higher copy number than nuclear DNA per cell, it is usually more difficult to obtain nuclear fragments than mt ones. Indeed, only one sample, CH559, allowed the amplification of 4 overlapping fragments of the IRBP gene, i.e. 271 bp. In summary, we were able to detect preserved DNA in 3 out of 4 samples analysed. Similar DNA preservation rates are reported in the literature for Canarian fossil specimens. These moderate success rates are probably due to the warm climatic conditions that prevail in the Canary Dabrafenib archipelago, known to be unfavourable for long-term DNA preservation. When targeting mt DNA, numt amplifications are common and largely documented for rodents. If not detected, numts can lead to an incorrect interpretation of the data and to an inaccurate phylogenetic position of the analysed species. In our study, overlapping fragments yielded an almost complete cytb sequence considered as the genuine mt copy. We also detected occasional amplifications of divergent sequences showing 5.4 to 20.9% of differences with this genuine sequence: a 97 pb fragment for CH476 and 5 fragments of 153 to 189 bp for CH559 for a total length of 402 bp were concerned. At first sight, these sequences were considered as numts since they were obtained at a very low frequency compared to the ����genuine���� ones. This observation is consistent with the fact that the number of mt genuine copies per cell is higher than the number of nuclear numt copies. However, we did not detect any stop codon or nucleotide insertion into these putative numts when translated into amino acids. The third position base composition of both kinds of fragments was also typical of the mammal cytb. Patterns of mutation rates at the codon positions were identical for all fragments considered. Both genuine and numt sequences, Nutlin-3 clinical trial although different, gave similar percentages of identity when compared by BLAST tool to sequences available in GenBank. Because of the slower mutation rate in nuclear DNA, numts may exhibit much shorter branch lengths when included in phylogenetic reconstructions and may be placed at a basal position in the tree.