Reading the molecular fingerprint of living systems — from single-cell gene expression to needle-free sensing through the skin.
Raman spectroscopy reads a sample's molecular vibrations directly, label-free and non-destructively. The lab pushes it in two directions: using machine learning to turn Raman spectra of living cells into single-cell gene-expression readouts, and miniaturizing Raman into compact sensors that read chemistry straight through the skin. Together they open biomedical applications across diabetes, aging, cancer, and image-guided surgery.
Two active fronts — one computational, one instrumental — each extending what a Raman spectrum can tell us about living systems.
Single-cell RNA sequencing reveals a cell's genetic program, but it destroys the cell — a snapshot, never a movie. Raman2RNA (R2R) infers genome-wide single-cell expression from label-free Raman microscopy instead. A full Raman spectrum is acquired at every pixel, and a neural network trained against single-molecule FISH and single-cell RNA-seq maps each living cell's spectral fingerprint into gene-expression space. Because Raman is non-perturbing, the same living cells can be read repeatedly over time: R2R followed expression dynamics through mouse embryonic-stem-cell differentiation and iPSC reprogramming, resolving where lineages diverge earlier than destructive snapshot methods could.
For people with diabetes, the standard glucose measurement is still a finger-stick blood draw — accurate, but painful and invasive enough that many test less often than they should. Minimally invasive continuous monitors (Dexcom, Abbott) have transformed daily management, but they still insert a microneedle filament under the skin, must be replaced every couple of weeks, and remain costly. A truly non-invasive optical reading could improve on both — if Raman can genuinely detect glucose through skin. The lab answered that question directly: with an off-axis fiber-optic Raman system it recorded the first direct observation of glucose's own Raman fingerprint in living tissue, tracking blood-glucose swings in vivo. Building on that proof, a band-pass Raman design reads just three narrow Raman bands — the glucose peak near 1125 cm⁻¹ and two references — shrinking the optics toward a compact, potentially wearable sensor.
Biomedical questions the lab takes on with its Raman imaging and sensing assets — spanning metabolic disease, aging, cancer, and the operating room.
