Anubis: a multi-level authentication scale for ancient proteins using random forest classification

Abstract

Deamidation is a time-dependent, post-translational modification with wide applications in clinical proteomics and palaeoproteomics as ageing indicators and authentication markers. Despite its significance, the use of deamidation for authenticating ancient proteins remains controversial, primarily due to variability in deamidation kinetics and context-specific protein preservation. We propose Anubis, an open-source, Python-based pipeline to evaluate relative deamidation patterns and authenticate ancient proteins. Anubis leverages random forest classification based on a range of tandem mass spectrometry-based features and physico-chemical characteristics. We use a combination of ancient and experimentally degraded beta-lactoglobulin to demonstrate that Anubis can authenticate a putative ancient dairy protein using position-specific deamidation patterns, relative deamidation (asparagine vs. glutamine), and overall modification patterns using trypsin processed in the same batch as a baseline. This multi-level approach addresses the complexities of ancient protein deamidation and paves the way for automated, reliable authentication across different archaeological substrates.