Forensic DNA phenotyping is the ability to predict an individuals physical appearance using genetic information obtained from biological material

This $980,527 federal Project Grant award from the National Institute of Justice (CFDA 16.560 - National Institute of Justice Research, Evaluation, and Development Project Grants) aims to enhance and expand the HirisPlex-S (HPS) forensic DNA phenotyping tool for predicting an individual's physical appearance traits such as eye, hair, and skin color.

The key objectives are to: 1) develop quantitative prediction models to complement the existing HPS web-based tool, 2) utilize deep learning to generate a database of images that can visually depict the genetically-inferred physical traits, and 3) expand the tool to include additional features like hair structure and eyebrow color. The enhanced HirisPlex-QT tool will be made available online to further the field of predictive biometrics. The project will also involve cultivating relationships with practitioners and researchers worldwide to discuss implementation and implications. The award period is from January 1, 2024 to December 31, 2026 and is being performed by Trustees of Indiana University.

Due to its intelligence-driven niche, IT has the potential to significantly complement current DNA analyses practice within forensic laboratories worldwide. However, in order to increase its usage and potentially move toward the implementation and use of these tools in current or cold casework, IT requires several developments that this proposal addresses. First, IT is time to enhance and expand the hirisplex-S (HPS) tool for eye, hair, and skin color prediction which has become a staple in the phenotyping field. Pigment is a continuous trait and up till now, we have utilized a simpler categorical classification (i.e., blue eyes, pale skin, blond hair) to predict the pigmentation appearance of an individual. The first aim of this proposal is to develop QUANTITATIVE prediction models that will complement the current HPS webtool available at https://hirisplex.erasmusmc.NL/ by systematically ranking all known associated pigmentation variants published in literature using a more globally represented dataset. Second, in order to produce realistic and genetically inferred visual predictions, as opposed to verbal descriptions of color, we shall utilize deep learning approaches to generate a database of images that can be used to relay the genetic prediction generated. By providing several image outputs, we remove the complications surrounding categorical definitions of color and interpretation bias that may be intrinsic to limited pigment descriptors. Third, due to the implementation of these visual predictions for eye, hair, and skin color, IT will be important to include additional traits and features that can relay the result more effectively but also have a solid genetic foundation. We will include the genetic prediction of hair structure (straight V not straight) and eyebrow color (QUANTITATIVE) to complement QUANTITATIVE hair and eye color visual predictions, as well as freckling (PRESENCE/ABSENCE) to complement the QUANTITATIVE skin color visual predictions. Finally, the enhanced tool termed hirisplex-QT and the expansion of QUANTITATIVE eye, eyebrow, hair, and skin color with freckling and hair structure (6 traits in total) shall be made available online in combination with the established HPS webtool to further the field of predictive biometrics. During its development, we aim to cultivate relationships with practitioners and fellow research laboratories around the world to discuss implementation and usage to ensure this transition proceeds with caution and appropriate guidance on the potential societal implications that surround visual physical appearance predictions. CA/NCF