Non-targeted metallomics through synchrotron radiation X-ray fluorescence with machine learning for cancer screening using blood samples

Abstract

Cancer is the leading cause of death in many countries. The development of new methods for early screening of cancers is highly desired. Targeted metallomics has been successfully applied in the screening of cancers through quantification of elements in the matrix, which is time consuming and requires combined techniques for the quantification due to the large elemental difference in the matrix. This work proposed a non-targeted metallomics (NTM) approach through synchrotron radiation based X-ray fluorescence (SRXRF) and machine learning algorithms (MLAs) for the screening of cancers. One hundred serum samples were collected from cancer patients who were confirmed by pathological examination with 100 matched serum samples from healthy volunteers. The serum samples were studied with SRXRF and the spectra from both groups were directly clarified through MLAs, which did not require the quantification of elements. The NTM approach through SRXRF and MLAs is fast (5s for data collection for one sample) and accurate (over 96% accuracy) for cancer screening. Besides, this approach can also identify the most affected elements in cancer samples like Ca, Zn and Ti as we found, which may shed lights on the drug development for cancer treatment. This NTM approach can also be applied through commercially available XRF instruments or ICP-TOF-MS with MLAs. It has the potential for the screening and prediction of other diseases like COVID-19 and neurodegenerative diseases in a high throughput and least invasive way.