Machine learning ATR-FTIR spectroscopy data for the screening of collagen for ZooMS analysis and mtDNA in archaeological bone

Faunal remains from archaeological sites allow for the identification of animal species that enables better understanding relationships between humans and animals, not only their morphological information, but also ancient biomolecules (lipids, proteins, DNA) preserved in these thousands even millions years. However, due to costs efforts required biomolecular analysis, there has been considerable research into development accurate efficient screening approaches remains. FTIR spectroscopy is one such approach considered its widespread use hindered by fact predictive accuracy can vary widely depending on extent sample preservation instrument used. Further, methods DNA (aDNA) analysis are scarce. Here we present a new vastly improve upon FTIR-based prior ZooMS (Zooarchaeology Mass Spectrometry) aDNA through random forest-based machine learning. To do so, ATR-FTIR examine three sets bone assemblages analyse them (for taxonomic identification). Two Palaeolithic contexts, dominated terrestrial fauna include specimens with variety preservational conditions. The third set consists Holocene faunal remains, variable levels cetaceans. Using were able more consistently evaluate ATR-FTIR-based mtDNA as well success. We report potential learning our models conclusively usage 20–40%. results suggest this potentially allows universal system, applicable across multiple largely independent spectrometers