Raman Spectroscopy and Machine Learning Reveals Early Tumor Microenvironmental Changes Induced by Immunotherapy.

Area of Science:

• Biomedical Optics
• Cancer Research
• Spectroscopy

Background:

• Identifying patients who benefit from cancer immunotherapy is challenging due to a lack of predictive biomarkers.
• Immune checkpoint inhibitors (ICIs) like anti-CTLA4 and anti-PD-L1 show durable benefit in a subset of patients.

Purpose of the Study:

• To apply label-free Raman spectroscopy for detecting biomolecular changes in the tumor microenvironment (TME) after ICI therapy.
• To develop spectral markers for predicting and evaluating response to anti-CTLA4 and anti-PD-L1 therapies.

Main Methods:

• Utilized label-free Raman spectroscopy on colorectal tumor xenografts treated with ICIs.
• Applied multivariate curve resolution-alternating least squares (MCR-ALS) for spectral decomposition.
• Employed support vector machine classifiers and random forests for response prediction and marker identification.

Main Results:

• Raman spectroscopy revealed early changes in lipid, nucleic acid, and collagen content post-therapy.
• Machine learning models achieved high prediction accuracy for ICI response.
• Identified therapy-specific spectral markers consistent with distinct ICI mechanisms of action.

Conclusions:

• Label-free Raman spectroscopy can sensitively detect early biomolecular alterations in tumors.
• These spectral changes predict response to cancer immunotherapy.
• This approach holds potential for clinical monitoring of ICI treatment efficacy.