The central challenge in modern diagnostics is not sensitivity — it is specificity at scale. Detecting a biomarker at picomolar concentration in a purified solution is a solved problem. Detecting the same signal in whole blood, cerebrospinal fluid, or tumour tissue, in the presence of thousands of competing molecular species, in a workflow compatible with a clinical laboratory — that remains a frontier.
My research addresses this challenge through the engineering of functional nanomaterials that are simultaneously sensing substrates and signal concentrators. The core platform is surface-enhanced Raman scattering (SERS): when a target analyte is adsorbed onto a plasmonic nanostructure, its Raman cross-section is amplified by up to 1010, enabling label-free, fingerprint-level molecular identification at trace concentrations. The critical advance my group has made is the integration of magnetic preconcentration — Fe₃O₄/Ag nanocomposites that can be directed to a target analyte in a complex matrix, magnetically separated, and interrogated by SERS — a strategy we call MA-SERS. This approach has demonstrated sub-nanomolar detection of dopamine in artificial cerebrospinal fluid and immunoglobulin G directly in blood, bypassing the sample preparation bottlenecks that have historically blocked SERS from clinical adoption.
The second axis extends the nanomaterial palette to two-dimensional structures — graphene, phosphorene, and their derivatives. These materials offer a chemically tunable surface, quantum-mechanical enhancement of Raman signals through GERS, and the structural flexibility required for drug loading and controlled release. Within the STRIKE Horizon Europe MSCA Doctoral Network, my group is developing 2D magnetic nanoplatforms that combine theranostic capability: the same nanostructure that delivers doxorubicin to a tumour cell can be tracked by SERS in real time.
Throughout both lines of work, data analysis is a co-designed component. Chemometric pipelines in Python and R, including multivariate classification and machine-learning-assisted spectral assignment, are integral to extracting clinical decisions from SERS datasets that are inherently high-dimensional and matrix-sensitive. The Raman chemical maps of tumour cells that my group produces are not images; they are spatially resolved molecular assays.
