Raman nanoparticle probes are an emerging new class of optical labels for interrogation of physiological and pathological processes in bioassays, cells, and tissues. Although their unique emission signatures are ideal for multiplexing, the full potential of these probes has not been realized because conventional analysis methods are inadequate. We report a novel spectral fitting method that exploits the entire spectral signature to quantitatively extract individual probe signals from multiplex spectra. We evaluate the method in a series of multiplex assays using unconjugated and antibody-conjugated composite organic-inorganic nanoparticles (COINs). Results show sensitive multiplex detection of small signals (<2% of total signal) and similar detection limits in corresponding 4-plex and singlet plate binding assays. In a triplex assay on formalin-fixed human prostate tissue, two antibody-conjugated COINs and a conventional fluorophore are used to image expression of prostate-specific antigen, cytokeratin-18, and DNA. The spectral analysis method effectively removes tissue autofluorescence and other unknown background, allowing accurate and reproducible imaging (area under ROC curve 0.89 +/- 0.03) at subcellular spatial resolution. In all assay systems, the error attributable to spectral analysis constitutes <or=2% of total signal. The spectral fitting method provides (1) quantification of signals from multiplex spectra with overlapping peaks, (2) robust spot-by-spot removal of unknown background, (3) the opportunity to quantitatively assess the analysis error, (4) elimination of operator bias, and (5) simple automation appropriate for high-throughput analysis. The simple implementation and universal applicability of this approach significantly expands the potential of Raman probes for quantitative in vivo and ex vivo multiplex analysis.