Defense Date


Document Type

Directed Research Project

First Advisor

Tal SImmons

Second Advisor

Baneshwar Singh

Third Advisor

Jenise Swall

Fourth Advisor

Claire Cartozzo


Estimating the post-mortem submersion interval (PMSI) can provide a valuable forensic tool for medicolegal death investigations involving victims discovered in aquatic environments. Previous studies conducted by Cartozzo et al. (2021) successfully demonstrated the use of microbial succession to create predictive models for the estimation of PMSI from submerged bone. Though effective, bone sampling requires time consuming processing techniques that result in destruction of decedent tissue. This study investigates the use of bone surface swabbing as an effective alternative method to bone sampling, with the goal of predicting PMSI using a simpler, non-invasive sampling technique. Porcine (Sus scrofa) skeletal remains (rib and scapulae) were caged and submerged in the James River at the Rice Rivers Center in Charles City, Virginia. One cage, containing five scapulae and five ribs, was collected every 250 ADD along with water samples (Cartozzo et al. 2021). In this study, swabs and water from the original experiment were analyzed at 500 ADD intervals, from baseline (0 ADD/0 days) to 4500 ADD (276 days). DNA was extracted from the swabs using the ChargeSwitch® gDNA Plant Kit and protocol, and variable region 4 (V4) of 16S rDNA was amplified and sequenced using the Illumina MiSeq Sequencing platform. Sequence analysis was performed with the Mothur (v.1.39.5) bioinformatics pipeline using the Mothur MiSeq SOP and R (v.4.1.1). Alpha diversity increased over the course of the study and Analysis of Molecular Variance (AMOVA) detected significant differences in beta diversity among bone, swab, and water groups (p<0.001, F=6.32137). These differences in beta diversity are likely explained by greater abundances of Clostridia and Gammaproteobacteria found in the bone samples compared to the swabs, and the overall variable presence and abundance of top taxa between bone and swab samples. Random forest models to predict PMSI were constructed using swabs for both ribs (R

2=0.822 and RMSE=600.6 ADD vs. R2=0.94, RMSE=477 ADD in

bone) as well as scapulae (R

2=0.766 and RMSE=681.4 ADD vs. R2=0.93, RMSE=501 ADD in bone). Swab samples predicted PMSI, albeit less accurately than bone powder, though this may well be due to the reduced sample of swabs (n=34) used in this study compared to bone (n=54). These results suggest that further investigation into bone surface swabbing is warranted, as improved models may provide an accurate, less labor-intensive, and non-destructive alternative for sampling skeletal remains to perform microbial PMSI prediction.


© The Author(s)

Is Part Of

VCU Master of Science in Forensic Science Directed Research Projects

Date of Submission