Performance limitations of label-free sensors in molecular diagnosis using complex samples

Label-free biosensors promised a paradigm involving direct detection of biomarkers from complex samples such as serum without requiring multistep sample processing typical of labelled methods such as ELISA or immunofluorescence assays. Label-free sensors have witnessed decades of development with a veritable zoo of techniques available today exploiting a multitude of physical effects. It is appropriate now to critically assess whether label-free technologies have succeeded in delivering their promise with respect to diagnostic applications, particularly, ambitious goals such as early cancer detection using serum biomarkers, which require low limits of detection (LoD). Comparison of nearly 120 limits of detection (LoD) values reported by labelled and label-free sensing approaches over a wide range of detection techniques and target molecules in serum revealed that labeled techniques achieve 2-3 orders of magnitude better LoDs. Data from experiments where labelled and label-free assays were performed simultaneously using the same assay parameters also confirm that the LoD achieved by labelled techniques is 2 to 3 orders of magnitude better than that by label-free techniques. Furthermore, label-free techniques required significant signal amplification, for e.g. using nanoparticle conjugated secondary antibodies, to achieve LoDs comparable to labelled methods substantially deviating from the original “direct detection” paradigm. This finding has important implications on the practical limits of applying label-free detection methods for molecular diagnosis.