Raman Spectroscopy and Machine Learning Team Up to Predict Immunotherapy Response in Patients

Using Raman spectroscopy and machine learning, a team at Johns Hopkins University developed a noninvasive technique to assess how cancer patients will respond to immunotherapy. The researchers used Raman spectroscopy to map the biochemical composition of tumors in detail, and machine learning to determine biomarkers indicating patient response to immunotherapy treatment. The Johns Hopkins team is the first to use label-free Raman spectroscopy to understand the biomolecular changes induced by immune checkpoint inhibitors in the tumor microenvironment.

Immunotherapy, a treatment that engages the immune system to fight cancer, helps only a fraction of patients. “Immunotherapy really works like magic and has fundamentally changed the way we view how cancer can be managed,” professor Ishan Barman said. “However, only around 25% of patients derive benefit from it, so there’s an urgent need to identify predictive biomarkers to determine who should receive the treatment.”

A noninvasive technique using an optical probe and Raman spectroscopy provides early signs of how a tumor is responding to immunotherapy. Courtesy of Adobe Stock.

Raman spectroscopy provides a precise molecular signature and, at the same time, is well suited for exploring the compositional changes of the tumor microenvironment and not just the cancer cells. “Rather than homing in on a few suspected molecules, we’re interested in getting a more holistic picture of the tumor microenvironment,” Barman said. “The tumor is not just the malignant cell. The microenvironment contains a complex combination of the tumor stroma, blood vessels, infiltrating inflammatory cells, and a variety of associated tissue cells.” The researchers used Raman spectroscopy to probe colorectal cancer tumors in three groups of mice. The first two groups received the two types of immune checkpoint inhibitors used in immunotherapy, and a third control group was left untreated.

Analyses of the Raman spectral data sets revealed early changes in lipid, nucleic acid, and collagen content following therapy. The team used the Raman data — approximately 7500 spectral data points from 25 tumors — to train an algorithm to determine a range of features caused by immunotherapy. “Our question was, can we differentiate between the three groups, and then what are the specific spectral features that are allowing us to differentiate between them,” Barman said. Using data from different mice, the researchers built a machine learning classifier and tested its performance. The goal was to construct a classifier that would mimic the biological variability the algorithm would encounter when presented with new data. “You need to prove beyond a doubt that the differences that you’re seeing are immune checkpoint inhibitor-induced, as opposed to just differences between two individuals,” Barman said.

The machine classifiers provided accurate predictions for responses to both immune checkpoint inhibitors and delineated spectral markers specific to each therapy. Although subtle, the differences in response were statistically significant, and this was corroborated by proteomics analyses conducted on the samples. The team’s observations of biomolecular changes in the tumor microenvironment, using Raman spectroscopy and machine learning, could inspire more detailed investigations into biomarkers and the further use of label-free Raman spectroscopy for clinical monitoring of immunotherapy response in cancer patients. Although more research is needed, the team believes its work could be used to develop a method for predicting whether a patient will respond positively to immunotherapy.

Raman spectroscopy has only recently been optimized for biomedical applications. “This is the first study that shows the ability of this optical technique to identify early response or resistance to immunotherapy,” researcher Santosh Paidi said. “Combined with machine learning, Raman spectroscopy has the potential to transform clinical methods for predicting therapy response.”

The research was published in Cancer Research (www.doi.org/10.1158/0008-5472.CAN-21-1438).