Rajas Poorna

BioE Ph.D. Proposal Presentation

2:00 PM, Monday, March 25, 2024

Location: EBB Krone – 4029 Conference Room 

Advisors:

Saad Bhamla, Ph.D. (ChBE, Georgia Institute of Technology)

Marcus Cicerone, Ph.D. (Chemistry & Biochemistry, Georgia Institute of Technology)

Committee:

Nicholas Hud, Ph.D. (Chemistry & Biochemistry, Georgia Institute of Technology)

Mark Prausnitz, Ph.D. (ChBE, Georgia Institute of Technology)

Francisco E Robles, Ph.D. (BME, Georgia Institute of Technology)

Bringing Universal Diagnostics to the Point-of-Care: Raman Spectroscopy, Sample Preservation, and Machine Learning

Raman spectroscopy (RS) of biofluids such as blood serum, urine, and saliva can diagnose a wide range of diseases such as diabetes, malaria, tuberculosis, celiac disease, cancer, and Alzheimer's. This has been shown on conventional Raman spectrometers costing >$100,000. We are developing a sub-$100 “frugal” Raman spectrometer with comparable sensitivity for broad-based biofluid diagnostics at the point-of-care (POC). We predict that our method will have sufficient signal for many diagnostics in just a 3-minute scan. Unlike Surface Enhanced Raman Spectroscopy (RS), our technique requires no special reagents, allowing the cost per test to drop to almost zero. We will first target tuberculosis (TB), for which such a fast, affordable test, deployed at TB centers, can save nearly 200,000 lives per year by reducing patient loss to follow-up.

In parallel, we are developing a fast, sub-$100 POC sample-drying instrument that enables biofluids to be preserved and transported without a cold-chain. This will enable remote POCs to offer a wider array of diagnostic tests without forcing the patient to travel. Our $60 prototype can dry a 1 mL sample in under 20 minutes where the Labconco Centrivap, a >$20,000 commercial instrument, takes over 5 hours (18x faster).

Finally, we have developed two machine learning techniques for analyzing the metabolomic cell state in Raman microscopy images. We use these to show that Raman microscopy can potentially identify the state of live cells to comparable or better resolution than transcriptomics. One of the techniques, SampleMAP, is of interest to the wider data science community as a dimensionality reduction technique with a significantly enhanced ability to identify clusters in data compared to UMAP.