Raman analysis of breast cancer-associated adipocytes: A chemometric pipeline for lipid biochemistry profiling

This study describes an integrated chemometric pipeline to analyse Raman spectra from breast tissue adipocytes, distinguishing Cancer associated adipocytes (CAAs) from normal adipocytes (NAs) and assessing the impact of obesity. Raman spectra were acquired from NAs and CAAs from the invasive front of breast tumor in 10 patients (5 normal weight, NW; 5 obese weight, OW). Extended Multiplicative Scatter Correction (EMSC) was adapted to correct carotenoid spectral interference. Random forest (RF) classifier was used for identifying discriminant wavenumbers and Uniform Manifold Approximation and Projection (UMAP) for visualization, with clustering quality assessed using silhouette scores. The results show the effectiveness of the pipeline in correcting the interferences and in identifying the key discriminant spectral regions. Informative wavenumbers highlighted differences in lipid unsaturation (C=C stretch at 1,655 cm^-1, =C-H stretching at 3,010 cm^-1), triglyceride composition (C=O stretching at 1745 cm^-1) and chain packing (CH₂ stretching 2,840-2,880 cm^-1), revealing greater biochemical heterogeneity in CAAs. In summary, this integrative approach of data processing and analysing provides an effective framework for studying subtle spectral differences in samples. The pipeline successfully distinguished CAA and NA phenotypes, establishing a foundation for identifying spectroscopic biomarkers of adipocyte pathological remodelling in breast cancer.