Departmental News

Posted:  Mar 24, 2025 - 01:00pm

Dr. Heather Edgar has received two awards for her significant accomplishments in the field of Dental Anthropology.  The Ellis R. Kerley Award Research Award  was presented at the meetings of the American Academy of Forensic Sciences. This award is presented for the best research presented by a member of the Anthropology Section of the Academy (see abstract below).  She also received the Innovation in Dental Anthropology Award at the Dental Anthropology Association meeting, which is held in conjunction with the American Association of Biological Anthropologists. This is the second year the award has been given. This award aims to recognize the contributions of a career scholar in dental anthropology. The innovation must have occurred within the immediate 10 years preceding the nomination for the award. Individuals will be recognized for pushing the discipline in exciting, novel, and ethical directions, which can include teaching, research, outreach, or service.

Introducing FoRDent for Estimating Population Affinity Using Dental Morphology

Heather Edgar, PhD, D-ABFA; Joseph Hefner, PhD, D-ABFA, Ron Richardson, BS

After attending this presentation, attendees will be aware of the availability, strengths, and capabilities of a new tool for estimating population affinity in forensic anthropological casework. They can apply this tool to improve the biological profiles produced toward resolutions for cases of unidentified human remains. The impacts of this presentation will include improving the diversity of data sources available for estimating population affinity, by providing practitioners with a new, freely available, statistically sophisticated, user-friendly method that is built on samples appropriate for forensic cases and other analyses. This method leverages underutilized data from the most taphonomically-resilient parts of the body, the teeth, to address one of the most complex components of the biological profile, population affinity.

FoRDent is a new tool for estimating population affinity from the permanent dentition. It is web-based and free access. Its interface is user-friendly, simple to use, and rich with images to facilitate use by practitioners who may not be experts in dental morphology. This simple to use interface is coupled with a large sample (n=4600) appropriate for forensic casework in North America, cutting edge statistics, and an approach to population affinity informed by contemporary anthropological thinking.

Dental morphology is an excellent data source for the estimation of population affinity in forensic casework. Teeth are hard, so they are often retained in cases even when taphonomic damage is significant to the skeleton. Dental morphological traits are heritable and generally neutral, so patterns of characteristics reflect biogeographic ancestry. These threshold characteristics are often scored categorically; however, traditional statistical methods for their analysis have required dichotomization of trait scores.

FoRDent draws on a near-global sample representing 111 modern nations, focused on the US and Mexico. US samples include recent and contemporary African, Asian, European, Hispanic, and Native Americans from 11 different states in different regions of the country. Additional contemporary samples represent Australia, Japan, and Mexico City, Tlaxcala, and Yucatán in Mexico. For contemporary samples, population affinities are as described by members of each local community from the samples were drawn.

Up to 65 dental variants can be scored and included in the analysis. All data are analyzed using the original (non-dichotomized) scores, preserving subtleties of variation and allowing for more fine-grained analyses. FoRDent employs a Gradient Boosting Machine (GBM) approach. GBM is an ensemble learning technique that builds a number of decision trees sequentially, with each tree correcting the errors made by previous models and boosting the final prediction as the sum of the predictions across all trees. GBM is highly effective classifying this sample, robustly handling this complex, multidimensional data. The FoRDent approach showcases the power of machine learning techniques in dental research and underscores their potential for dental morphology in population affinity estimates going beyond more naïve approaches that do not weigh variable importance. Users can refine analyses through variable selection approaches emphasizing these measures. Further, users are provided multi-level results, with probabilities associated with continental, national, and community level affinities. Classification accuracies vary depending on the variables and samples included in an individual analyses. Including all groups and all traits, which provides the lowest estimate of accuracy, still correctly classified nearly 50% of the sample. A more typical analysis comparing an unknown individual to 4 regional samples, results in classification accuracies between 83.9 and 91.1 percent.