Artificial intelligence (AI) is used increasingly to process and analyze data collected through Raman spectroscopy. Chemometric techniques, which are data-driven and include spectral processing and statistical analysis, can be used to detect and extract information from subtle differences in Raman spectra. However, established standards for Raman spectral analysis have yet to exist. To maximize the value of AI analysis for spectral data, a research team from the Leibniz Institute for Photonic Technologies (Leibniz IPHT) and Friedrich Schiller University developed a guide for Raman spectral analysis. The guide covers each step in the analytical process, from experimental design to statistical analysis, and provides strategies for circumventing issues. The standards will make it easier for scientists to use Raman spectroscopy for real-world applications in biology, diagnostics, food safety, pharmaceuticals, and other fields. “In order to help Raman spectroscopy achieve a breakthrough in applications, standardized workflows are needed that deliver results that are as robust as possible,” professor Thomas Bocklitz, head of the Photonic Data Science research departments at Leibniz IPHT and the University of Jena, said. Bocklitz worked on the project with professor Jürgen Popp, who leads the Spectroscopy/Imaging Department at Leibniz IPHT. Popp’s group contributed its expertise in Raman spectroscopy for the fields of medicine, life and environmental sciences, quality and process analytics, and pharmaceuticals. Bocklitz’s group works on the computer-aided evaluation of Raman spectra at the conceptual level. The guide developed by Bocklitz and Popp shows scientists how to perform a Raman spectral analysis using data-driven techniques, including machine learning-based modeling. It is divided into four parts: experimental design, data preprocessing, data learning, and model transfer. To illustrate the steps, the researchers depicted a standard workflow with three sample data sets where spectra from individual cells have been collected in single-cell mode, and with one sample data set where data has been collected from a raster scanning-based Raman spectral imaging experiment on mice tissue. A graphic shows a set of instructions for standardized Raman spectral analysis. A team of leading Raman spectroscopists aims to develop a set of standards that shows how to perform Raman spectral analysis using data-driven techniques. Courtesy of Thomas Bocklitz, Leibniz IPHT. When used with Raman spectroscopy, chemometric techniques can help scientists identify the spectral differences that are most useful for differentiating between cell types. Raman spectroscopy enables users to acquire complex measurement data including extensive molecular fingerprints. This data can be interpreted using AI. Further, by analyzing the molecular fingerprints of samples, scientists can characterize materials based on their chemical composition, identify pathogens, or detect diseased tissue. The signals and signal differences within the measurement data are minimal and are affected by multiple variables. AI techniques can be useful for sifting through the nuances in the data. Next, the team plans to work with additional research institutions to focus the protocol on instrument intercomparability, which it will do through a joint ring trial that will explore how to correct for instrument dependence. The team also plans to use the standardized methods for AI-based evaluation of Raman spectra to develop market-ready, light-based diagnostic methods and new therapeutic approaches at the future Leibniz Center for Photonics in Infection Research, in Jena. The standards introduced by the team could help shift Raman-based technologies from primarily a proof-of-concept tool to a method that is used more extensively for practical applications in health, environment, medicine, and other areas. The research was published in Nature Protocols (www.doi.org/10.1038/s41596-021-00620-3).