Proteomic signature of aging in bloodstain samples: a preliminary study

TL;DR

This study explores using proteins in bloodstains to estimate age, offering a promising alternative to DNA-based methods.

Contribution

The study is the first to use proteomic signatures from bloodstains for forensic age prediction.

Findings

Age-related proteins showed four nonlinear change patterns during aging.

The Boruta-based model achieved an R² of 0.70 and a MAE of 9.14 years.

Age-associated proteins were enriched in endocytosis, metabolism, and neurodegenerative disease pathways.

Abstract

Age estimation from biological samples remains a critical challenge in forensic investigations, particularly when analyzing trace or degraded biological stains recovered from crime scenes. While DNA methylation has gained attention as a potential epigenetic clock for age prediction, practical limitations that include the requirements for bisulfite conversion and environmental interference hinder its forensic utility. Previous studies have shown that proteins are more stable than DNA and that certain proteins are highly age-related and gender-differentiated, indicating that proteomic signatures present a promising alternative for biological age determination. However, proteomic data from bloodstains have not yet been used for forensic age prediction. In this pilot study, we used the high-resolution Thermo Scientific Orbitrap Astral Mass Spectrometer to investigate the proteomic signatures of aging in bloodstain samples from 40 healthy males (aged 10–79 years) stored for 4 years at room temperature (20–25 °C). Age-related proteins were subsequently selected for age prediction. We further simplified the characteristic variables using the Least Absolute Shrinkage and Selection Operator (Lasso) regression and the Boruta algorithm, and established age prediction models for bloodstains based on Random Forest (RF) machine learning.

In total, 1,655 proteins were identified, which showed four different nonlinear change patterns during the aging process. Pearson’s correlation coefficient (R) was calculated, and 71 proteins were found to correlate significantly with age (Pearson’s |R| > 0.3, P < 0.05), including 26 positively and 45 negatively correlated proteins. Functional enrichment analysis revealed that age-associated proteins were markedly enriched in pathways related to endocytosis, metabolism, and neurodegenerative disease. Feature selection using the Lasso regression and the Boruta algorithm resulted in the identification of 18 and 10 age-associated proteins, respectively, with six overlapping proteins (including ITIH3, HSPA9, SNAP91, FTL, XPO4, and NCF2). RF regression models were constructed using different feature sets: Lasso-selected (18 proteins), Boruta-selected (10 proteins), their intersection (6 proteins), and R-value-based filtering (e.g., top 7 proteins with |R| > 0.4, P < 0.05). The Boruta-based model demonstrated the highest predictive accuracy, achieving an R² of 0.70 and a Mean Absolute Error (MAE) of 9.14 years for the testing set.

These findings demonstrate the potential of bloodstain proteomics for age estimation. This study provides a foundation for further validation in larger cohorts and diverse forensic scenarios.

The online version contains supplementary material available at 10.1186/s12864-025-12164-x.

Figures