The use of non-invasive genetic samples to identify individual animals is becoming increasingly common in wildlife studies. Sampling DNA non-invasively (often from hair or faeces) has advantages for identifying animals for uses such as mark-recapture modelling that require unique identification of animals in samples. These advantages include removing the need to capture or handle animals that are rare or difficult to catch, such as pests or endangered species. The 'genetic tag' can be read without directly interacting with the animal. This method has huge potential, but while it is possible to generate significant amounts of data from these non-invasive sources of DNA, a challenge in the application of the approach is overcoming genotyping errors inherent in samples collected. These errors can lead to incorrect identification of individuals.
Genotyping errors can arise when poor sample quality due to an insufficient quantity of DNA leads to failure of DNA amplification at one or more loci. This has the effect of heterozygous individuals being scored as homozygotes at those loci as only one allele is detected (termed 'allelic drop-out'). These error rates will be specific to a species, and will depend on the source of samples and the way the samples have been handled.
If errors go undetected and the genotypes are naively used in mark-recapture models, significant overestimates of population size can occur. Incorrect genotypes lead to encounters of 'new' individuals and also cause a false decrease in the probability of recapture. Occasionally identical genotypes may arise, and these are treated as recaptures.
With good quality samples genotyping error is negligible. One common approach designed to avoid genotyping error is to reject low quality samples but this may lead to the elimination of large amounts of data. It is preferable to retain these substandard samples with low DNA concentration as they still contain usable information in the form of partial genotypes. Rather than trying to minimise error or discarding error-prone samples this research models the presence of genotyping error due to dropout in the analysis.
Using data from brush-tailed possum (Trichosurus vulpecula) in New Zealand and the Eurasian badger (Meles meles) in the United Kingdom, this thesis describes a method based on Bayesian imputation that allows us to model data from non-invasive genetic samples in the presence of genotyping error in the form of allelic dropout. This model makes it possible to include samples with uncertain genotypes (for example those with low sample DNA concentration or samples with information completely missing at one or more loci).
The urn sampling model (where all individuals are assumed to have an identical probability of capture) has been extended to allow for heterogeneity in capture probability (the vase sampling model). Model selection indicates that for the badger data this extension is necessary.
The model presented here could be extended in several ways. Elaborating the sampling model by incorporating covariates (e.g. sample location) may give more information for resolving ambiguity in identification. Other types of genotyping error (such as false alleles) could be incorporated into the model for genotype corruption if necessary for the data being analysed. It would be useful to extend to modelling open populations, both from the standpoint of obtaining better parameter estimates and also as a device for studying relationships among individuals. For the analysis presented here, inferring the underlying true genotypes for each sample was a byproduct of the analysis. In many situations (such as when studying pedigrees, disease association or plant and animal breeding experiments) these uncorrupted genotypes will be of great interest in their own right.
