AI Ramanomics method identifies organelles without dyes

University at Buffalo-led researchers have developed a label-free way to study living cells by combining AI with “Ramanomics,” an optical technology that measures cellular biochemistry without altering samples. The method replaces fluorescent dyes, which can disrupt natural cell behavior, limit how many structures can be examined at once and reduce measurement accuracy.

The team used Raman spectroscopy to capture biochemical “fingerprints” from four cellular organelles and trained several machine learning models to recognize them. A neural network performed best, correctly identifying organelles with about 90% accuracy and locating where new Raman measurements were collected without fluorescent labels.

Researchers say the approach could help track how diseases affect cells, monitor drug responses and identify molecular changes linked to cancer and metabolic disorders. The group is also working to pair its models with Raman imaging systems that use quantum light sources, aiming to shorten signal collection from up to three minutes to about a second and generate larger datasets for stronger models.