The throughput of spontaneous Raman spectroscopy techniques for cell identification is limited due to the relatively low sensitivity and long acquisition time. This throughput can be increased by surface enhanced Raman scattering (SERS). Common SERS approaches for cell identification suffer from reproducibility and complex nanoparticle preparation. Here we present a new strategy for identification and classification of cells using silver nanoparticles aggregates and cell lysate prepared by ultrasonication. The probe ultrasonic system disrupts the cell wall and enables interaction of released cell components to nanoparticles. This approach was applied to distinguish four cell lines – Capan-1, HepG2, Sk-Hep1 and MCF-7 – using SERS at 785 nm excitation. Six independent batches were prepared per cell line to check the reproducibility. Principal component analysis was applied for data reduction and assessment of spectral variations that were assigned to proteins, nucleotides and carbohydrates. Four principal components were selected as input for a classification model based on support vector machine which identified and classified the four cell lines with sensitivity, specificity and accuracy above 95%. The numbers were obtained by leave-three-batches-out cross validation. This proof-of-principle demonstrates a reproducible and highly specific SERS approach for cell identification with easy to prepare silver nanoparticles.