Fourier transform infrared spectroscopy in plasma cell dyscrasia testing

Abstract Objectives Plasma cell dyscrasias (PCDs), including monoclonal gammopathy of undetermined significance (MGUS) and multiple myeloma (MM), are characterised by clonal plasma cell proliferation producing monoclonal paraproteins. Diagnosis relies on laboratory-based assays with variable turnaround times. Fourier transform infrared spectroscopy with attenuated total reflectance (FTIR-ATR) offers rapid, non-destructive biochemical profiling and may support PCD testing. Methods We studied 25 PCD patients (10 MGUS, 15 MM) and 20 healthy controls. Plasma spectra were acquired between 4,000 and 800 cm −1 . Pre-processing included standard normal variate, generalised least squares weighting, and mean centering. Four multivariate models – principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), artificial neural network discriminant analysis (ANNDA), and linear discriminant analysis (LDA) – were trained and evaluated with cross-validation. The mean FTIR spectra of the PCD groups were deconvoluted in the glycation (950–1,480 cm −1 ) and lipid (2,800–3,600 cm −1 ) regions to identify overlapping peaks and the corresponding differences in specific molecular functional groups or structural features. Results All models achieved perfect calibration (true positive rate [TPR] 1.00, false positive rate [FPR] 0.00). In cross-validation, ANNDA achieved the highest accuracy (94.1 %), followed by LDA (91.1 %) and PLS-DA (88.9 %). Specificity for healthy controls exceeded 95 % in all models, with minimal false positives. MGUS–MM differentiation was less accurate (TPR 0.80–0.93), with misclassifications primarily between these groups. Deconvolution of the glycation and lipid regions of the FTIR spectra showed 11 statistically significant overlapping peaks. Conclusions FTIR-ATR spectroscopy with machine learning could differentiate healthy controls from PCD with high accuracy and specificity. Use of larger externally validated data sets would allow refinement of the models; nonetheless, this demonstrates the future diagnostic capability of FTIR in PCD with the potential to apply this technology to other target diseases.