*Corresponding Author: S. Bavankumar, Research Scholar, Annamalai University, Chidambaram, Tamil Nadu, India - 608002. Email: sbavankumar55@gmail.com
After mass casualty events and in the context of complicated forensic analysis, it is very important to quickly and accurately identify missing people for both legal and humanitarian reasons. Short Tandem Repeats (STRs) are the most reliable way to identify a person, but traditional STR analysis is slow and prone to mistakes because it relies on both manual and statistical analysis. This research is especially pertinent in the realm of extensive disasters and tainted DNA specimens.
This paper introduces a deep learning model based on Convolutional Neural Network (CNN) for the automated analysis of DNA patterns obtained from STR analysis. The STR allele data is transformed into a format that deep learning can use, allowing the network to learn and identify complex DNA patterns in large datasets. In accordance with accepted forensic analysis standards, the system is trained and evaluated on both simulated and actual STR analysis. CNN and CNN-Bidirectional models were developed using Label and K-mer encoding techniques to further assess classification accuracy. The tests showed that the K-mer encoding CNN and CNN-Bidirectional models achieved 98.99% accuracy.
The new system identifies patterns much faster and with significantly higher accuracy, making it useful for applications such as mass graves and mass casualty events in forensic science. This study improves the accuracy and effectiveness of DNA identification and lays the groundwork for future AI-based forensic analysis.
Keywords: DNA, CNN, STR, Mass Casualty Incidents.
How to cite this article: Bavankumar S, Rathikarani V, Santhoshkumar R. Autonomous DNA pattern intelligence for large-scale forensic human identification using deep learning technique. Int J Drug Deliv Technol. 2026;16(3s): 890-898; DOI: 10.25258/ijddt.16.3s.109
Source of support: Nil.
Conflict of interest: None