In a publication in Cancers a comprehensive mass spectrometry strategy was used to profile blood plasma samples from a large pan-cancer cohort including 15 different cancer types. Differential expression analysis and machine learning was then used to identify a potential biomarker panel able to correctly diagnose multiple myeloma.

A LC-MS/MS-based targeted proteomics strategy that has been shown to accurately measure protein concentrations in complex biological samples was used, in combination with stable isotope standard protein fragments to enhance quantitative performance, to analyze almost 2000 blood samples from cancer patients. For each of the samples an assay with more than 200 proteins, including FDA approved targets, complement cascade proteins and proteins annotated to be secreted in blood, was quantified in multiplex.

The results show molecular phenotypes that could distinguish patients with different cancers and also a specific decrease in complement C1 members in patients with multiple myeloma (MM). To further investigate the plasma proteome of MM patients a model based on the random forest algorithm was trained to predict disease outcome, and by combining the markers from machine learning and differential expression MM patients could be distinguished from all other cancer diagnoses with very high confidence.

Link to article