Cancer diagnoses traditionally require invasive or labor-intensive procedures such as tissue biopsies. Researchers at the Ludwig-Maximilians-Universität München (LMU) have now reported on a method that uses pulsed infrared light to identify molecular profiles in blood plasma that could indicate the presence of some common cancers. In a proof-of-concept study, the team used machine learning to analyze blood plasma from more than 2,000 participants and link molecular patterns to lung cancer, extrapolating a potential “cancer fingerprint.”
Mihaela Žigman, PhD, at the Max Planck Institute of Quantum Optics (MPQ) is corresponding author of the team’s published paper in ACS Central Science, titled “Electric-Field Molecular Fingerprinting to Probe Cancer.” In their report, the team concluded, “Our study demonstrates that electric-field molecular fingerprinting is a robust technological framework broadly applicable to disease phenotyping under real-world conditions.”
Various physiological states, and some diseases, may be reflected in the molecular makeup of biofluids such as blood and plasma, which is the liquid portion of blood depleted of any cells, the authors explained. Plasma carries diverse molecules such as proteins, metabolites, lipids, and salts throughout the body and some molecules carried by blood plasma indicate potential health conditions. For example, unusually high levels of prostate-specific antigen are used to screen for prostate cancer. Theoretically, a medical test that measures a broad range of molecules could identify a pattern specific to different cancers, leading to quicker diagnoses and reduced costs.
The use of sensitive and specific analytical methods in the fields of proteomics and metabolomics has led to the discovery of various molecular “biomarker candidates” the authors noted. “However, current omics techniques are often still limited in the range of molecular species that they can probe at once. They often require complex, target-specific preanalytical workflows for sample preparation.”
A technique known as molecular fingerprinting, “… where phenotype detection is based on patterns of change across the entire molecular landscape,” could represent an alternative approach, the researchers continued. “If a specific pattern shows a robust correlation with a particular physiological state, it may contribute to the detection of a phenotype.”
For their newly reported study, Žigman and colleagues tested a technique called electric-field molecular fingerprinting (EMF), which uses pulsed infrared light, to profile complex molecular mixtures in blood plasma and look for telltale chemical patterns of cancer.
First, the researchers used the electric-field molecular fingerprinting technique to send ultra-short bursts of infrared light through plasma. They analyzed samples from 2,533 study participants, including people with lung, prostate, breast, or bladder cancer and those without cancer. For each sample, they recorded the pattern of light emitted by the molecular mixtures in the plasma, the “infrared molecular fingerprint.” Using these complex patterns from individuals with and without cancer, the researchers taught a machine learning model to identify molecular signatures associated with the four types of cancer.
The computer model was tested on a separate subset of participants’ samples to see how well the model could perform on unseen test data. The analytical technique demonstrated a convincing level of accuracy (up to 81%) in detecting lung cancer-specific infrared signatures and differentiating them from control samples obtained from individuals without cancer. “In an independent held-out test data set, designed to reflect different experimental conditions from those used during model training, we achieved a lung cancer detection ROC AUC of 0.81… This independent testing allows us to assess the generalizability of our technique beyond a single measurement campaign, revealing that our lung cancer detection model remains robust under realistic measurement shifts, maintaining its diagnostic performance and demonstrating its potential reliability,” they noted.
The computer model’s performance did demonstrate lower success rates in detecting the other three cancers. The observed discrepancies, particularly in the capacity to detect the other three cancer entities, the authors noted, “… highlight the need for improving the reproducibility of EMF measurements and for further validation of this approach in additional patient populations.”
Nevertheless, they wrote, “Our findings indicate that patterns in infrared fingerprints can reliably be associated with physiological states … In conclusion, the current findings provide compelling evidence underscoring the potential of electric-field molecular fingerprinting for minimally invasive disease detection.”
In the future, the team aims to expand and test the approach to identify additional cancer types and other health conditions. “Future enhancements, such as broader spectral coverage, increased detection sensitivity and specificity, multidimensional measurements, and interferometric subtraction, could further boost biomedical potential,” they stated. “Expanding clinical studies to larger cohorts, focusing on early disease states and independent clinical testing, and exploring various disease phenotypes and their combinations will be crucial for developing a reliable diagnostic platform to improve cancer outcomes.”
Žigman added, “Laser-based infrared molecular fingerprinting detects cancer, demonstrating its potential for clinical diagnostics. With further technological developments and independent validation in sufficiently powered clinical studies, it could establish generalizable applications and translate into clinical practice—advancing the way we diagnose and screen for cancer today.”