The performance of deconvolution across different Short Tandem Repeat (STR) detection kits often degrades due to variations in fluorescence labeling systems and allelic ladders. Furthermore, the scarcity of samples from target kits frequently leads to model overfitting. To address these challenges, this study proposes a few-shot cross-kit intelligent deconvolution method for mixed STR profiles based on Domain Adaptation (DA). This method aims to extract generic STR profile features by aligning data distributions across different kits. First, a deep feature mapping network is constructed, incorporating a domain alignment constraint based on Maximum Mean Discrepancy (MMD). This minimizes the discrepancies between source and target kits regarding STR peak shape distributions and baseline noise, thereby extracting domain-shared features that are independent of specific kits. Second, the model undergoes joint training by simulating few-shot cross-domain tasks on multi-source kit data. This process encourages the model to overcome feature bias towards any single kit and learn generic parameter representations with high generalization capability. Finally, by integrating a kit-aware attention mechanism, the method performs efficient fine-tuning with only a minimal number of target kit samples, dynamically adapting to the specific fluorescence characteristics of the target kit. Experimental results demonstrate that this method effectively overcomes the Domain Shift phenomenon and maintains high deconvolution accuracy even when target kit samples are extremely limited. These findings provide a practical technical solution for the cross-platform intelligent analysis of forensic DNA mixed STR profiles.

Authors

• Liang Zeng (Beijing Institute of Technology, School of Cyberspace Science And Technology, China)
• Yupeng Hu (Beijing Institute of Technology, School of Cyberspace Science And Technology, China)
• Zhen Xu (Institute of Forensic Science, Ministry of Public Security, China)
• Fan Yang (Institute of Forensic Science, Ministry of Public Security, China)