Rationale: Pre-screening of bone collagen quality is important to reduce the cost for analyses such as radiocarbon (14 C) dating with accelerator mass spectrometry in archaeological studies. We developed a pre-screening protocol based on attenuated total reflection (ATR) Fourier-transform infrared spectroscopy (FTIR) for assessing the chemical composition and mineralogy of ancient bone samples.
Methods: ATR-FTIR measurements were performed on bulk bones of diverse origin and age before collagen extraction. The percentage nitrogen of bulk bones, as well as the weight percentage, and the percentage carbon and nitrogen of extracted organic matter were noted. Several machine learning algorithms were applied to the spectral data and compared for their efficacy in screening for well preserved collagen.
Results: The results showed that (a) the first derivative of the spectral data was better suited to screening than the raw FTIR data, especially for a wider spectral range and (b) certain classification algorithms [e.g. gradient boosting machine (GBM)] were able to efficiently predict the degree of preservation in bone samples.
Conclusions: This pre-screening protocol can serve as a fast, concise and inexpensive pre-screening tool for determining relative degrees of preservation before collagen extraction and subsequent 14 C dating. The screening power based on the machine learning techniques can be further improved by accumulating the FTIR spectral data of bones.
© 2020 John Wiley & Sons, Ltd.