Given the relative difficulty in obtaining well-preserved samples and extracting DNA from fossil bone, all PCR reactions were done in relatively small volumes, using just 1 ml of the undiluted DNA PCI-32765 Src-bcr-Abl inhibitor extract per reaction. PCR products were screened with electrophoresis on 2% agarose gels. If bands were visualised in the expected size-range, primers were re-ordered with a 59 fluorescent 6-FAM dye and the PCRs were repeated with these. DNA fragments were separated on an ABI 3730 genetic analyser and sized with Genescan LIZ500 size standard. Negative and positive controls were always included. The alleles were scored using GENEMARKER version 1.5. From this study, we have shown that it is possible to generate a fully functional microsatellite library from LCN aDNA templates. Several studies have amplified microsatellite markers from more recent museum specimens, for example to investigate a loss of genetic biodiversity resulting from recent environmental changes e.g.,. Here, we show that reliable microsatellite data can be generated from samples of much greater antiquity, but that care needs to be exercised in compiling the data. It is clear from the data that allele scoring was seriously hampered by allelic dropout. Indeed, if the GDC-0449 common approach with just one PCR per marker was applied, an overall average of 53% of the heterozygotes would likely have been misidentified as homozygotes. Likewise, attempts to multiplex even two of these markers together caused even greater dropout and cannot be advised for templates such as these. As mentioned above, the problem is well described in the literature although the average dropout we observed seems slightly worse than in studies of, for example, faecal DNA: 24% dropout in Morin et al. and 29% dropout in Frantz et al., DNA from 100-year-old teeth: 42% dropout in Arandjelovic et al., shed hair: 31% dropout in Gagneux et al., and human fingerprints: 12% in Balogh et al.. The three criteria, proposed to minimise the effect of dropout on our moa data are discussed below: Criterion 1 involved a 4x singleplex PCR repeat method for each apparent homozygote. According to the average error rates shown in Table 3, this should reduce genotyping errors to affect between 0.3% and 5.9% of the heterozygotes per marker. When summed, a theoretical maximum of 6.6% of all genotyped individuals in the final data would contain an error. This 4x approach was based on balancing the generation of high quality data, against maintaining enough DNA extract to amplify all markers. The evaluation had to factor in the relative difficulty in obtaining sample material for additional extractions, and also account for a high rate of PCR failures, given that retrieving results from four positive PCR reactions could for some extracts easily involve 6�C10 PCR setups. Our approach here is less conservative than the often cited 7x PCR repeat suggested by Taberlet et al., which was based on computer simulations of a worst-case scenario with 100% allelic dropout.