Automated separation of overlapping fingermarks by non-negative matrix factorization of DESI mass spectrometry imaging data.

Overlapping fingermarks present a significant challenge in forensic investigations, as traditional optical methods cannot separate superimposed ridge patterns from multiple donors. We present an automated and unsupervised approach using non-negative matrix factorization (NMF) to separate overlapping fingermarks from desorption electrospray ionization mass spectrometry imaging (DESI-MSI) data by exploiting the distinct chemical signatures of individual donors. DESI-MSI datasets containing 100-200,000 pixels with intensity measurements up to 5000 mass-to-charge channels were decomposed using NMF to identify spatially coherent molecular signatures. Using PCA as a baseline, NMF produced superior results with sharper spatial boundaries between fingermark regions, directly interpretable non-negative molecular signatures, and parts-based decomposition that naturally represents the additive nature of overlapping chemical deposits. The completely unsupervised analysis successfully identified distinct fingermark donors and their associated molecular profiles without requiring prior knowledge of donor number or chemical characteristics. The approach also successfully resolved a simulated four-donor scenario, demonstrating scalability beyond two overlapping fingermarks. We recommend NMF as the preferred method for detailed forensic fingermark analysis, since it offers forensic practitioners an automated, interpretable tool for analyzing complex fingermark evidence where multiple deposits overlap. The approach is likely to extend to other MS imaging areas such as document analysis.