Data Analytics, Statistics, Chemometrics, and Artificial Intelligence

A recent review article explores how machine learning (ML)-assisted Raman spectral classification is being used in applications such as biomedicine and material analysis.

Artificial intelligence and machine learning are rapidly reshaping how analytical data are modeled, interpreted, and deployed, but the conceptual foundation is already familiar to practitioners of chemometrics. Latent variables, calibration models, variance–bias tradeoffs, and multivariate optimization did not originate with neural networks; they have been central to spectroscopic data analysis for decades. This expanded glossary provides a rigorous, side-by-side translation between modern artificial intelligence (AI) terminology and established chemometric concepts. This glossary is intended to demystify AI terminology, while preserving statistical clarity. It is designed to help analytical scientists, spectroscopists, and chemometricians engage with modern data-driven methods without abandoning physical interpretability or statistical discipline.

Spectroscopy is seeing several important changes in the industry. In this feature, we focus on four topics that have emerged.

A recent study demonstrates that updated predictive models based on NIR spectra can outperform traditional nitrogen-based prescreening methods in identifying samples suitable for radiocarbon dating.

From a chemometric standpoint, artificial intelligence (AI) in spectroscopy is best understood as an extension of established multivariate methods rather than as a replacement. Most AI approaches closely parallel familiar tools such as regression, classification, and principal component analysis, but offer greater flexibility to handle nonlinear behavior, interacting physical and chemical effects, and large, heterogeneous datasets. By learning directly from raw spectra, AI methods can reduce reliance on manual preprocessing while still indicating which spectral regions influence predictions. In this sense, AI represents a developmental layer of chemometrics that enables classical concepts to operate effectively in modern spectroscopic systems. Overall, AI is best viewed as the next developmental layer of chemometrics, not as a competing discipline. As with all current AI programs, domain knowledge of analytical chemistry is essential for AI’s effective application. Knowing the boundaries of what is plausible in any chemical or modeling system allows fine-tuning of the models towards useful and reliable analytical results.

At Pittcon, generative artificial intelligence will be presented at the James L Waters Symposium on Monday, March 9, 2:30 PM to 4:40 PM in Room 221A. Generative artificial intelligence has transitioned from a conceptual novelty to a practical approach for innovation in spectroscopic data analysis. During 2025, a small set of highly influential publications crystallized this transformation by demonstrating how generative models can synthesize realistic spectra, solve inverse spectral problems, accelerate materials discovery, and automate molecular structural elucidation. This article reviews six pivotal contributions published in 2025 that collectively define the state of generative artificial intelligence in spectroscopy. These works establish theoretical foundations, survey emerging methods, introduce physics-informed generative architectures, and demonstrate transformative applications across vibrational, electronic, and magnetic resonance spectroscopies.

Spectroscopists are routinely embracing complex algorithms to handle high-dimensional, nonlinear, and multimodal data. Why is this happening? We explore this question in this tutorial.

During 2025, near-infrared (NIR) spectroscopy has accelerated its transition from a mature analytical technique into a digitally enabled cornerstone of biopharmaceutical manufacturing and quality control. Advances in miniaturized instrumentation, process analytical technology (PAT), chemometrics, artificial intelligence (AI), and real-time process control technologies have driven NIR spectroscopy into new roles spanning upstream fermentation, downstream processing, raw material characterization, and continuous manufacturing. This article reviews and contextualizes ten influential peer-reviewed publications from 2025 that collectively define the current state and near-term trajectory of NIR spectroscopy in biopharmaceutical analysis.

Fourier transform infrared (FT-IR) spectroscopy has undergone a notable evolution in biopharmaceutical analysis over the past three years. Advances in crystal engineering, process analytical technology (PAT), chemometrics, machine learning (ML), and hyphenated analytical platforms have significantly expanded FT-IR’s analytical scope. This article reviews ten of the most influential publications from 2023–2026 that exemplify FT-IR’s growing role across the biopharmaceutical lifecycle, from drug substance design and formulation to manufacturing, quality control, and clinical bioanalysis.

Between 2023 and 2026, Raman spectroscopy transitioned from a supportive analytical technique to a central enabling technology in biopharmaceutical analysis and manufacturing. Advances in artificial intelligence (AI), machine learning (ML), automation, and surface-enhanced Raman spectroscopy (SERS) have expanded Raman’s role from nutrient monitoring to real-time prediction of critical quality attributes (CQAs), inline control of complex bioprocesses, and non-destructive analysis of finished drug products. This article reviews ten of the most influential publications from this period, highlighting how they collectively reshaped expectations for Raman spectroscopy as a process analytical technology (PAT) and a quality-by-design (QbD) tool in modern biopharmaceutical development.

Over the past two years, near infrared spectroscopy (NIRS) and related NIR techniques have seen rapid adoption in biomedical research. These developments span non invasive diagnostics, functional monitoring, machine learning integration, point of care probes, and applications in complex clinical settings such as liver fibrosis, viral detection, neonatal care, brain injury, and neurodegenerative disorders. This article synthesizes 10 key publications, highlighting trends, methodologies, and clinical potential.

A recent study demonstrated that combining laser-induced breakdown spectroscopy (LIBS) with machine learning (ML) can accurately identify gunshot residue from nontoxic ammunition and reliably distinguish shooters from non-shooters despite the absence of traditional elemental markers.

Over the past three to four years, Fourier Transform Infrared (FT-IR) spectroscopy has emerged as one of the most rapidly expanding vibrational techniques in biomedical research. Driven by advances in attenuated total reflectance (ATR), live-cell measurements, chemometrics, and machine learning (ML), FT-IR has moved beyond descriptive biochemical profiling toward predictive diagnostics and translational clinical science. This article highlights and critically summarizes the top 10 most influential peer-reviewed articles published recently on FT-IR applications in tissues, cells, hair, blood, saliva, urine, and exercise physiology, emphasizing analytical innovation, clinical relevance, and future impact.

In the past few years, Raman spectroscopy and its technological modifications—such as surface-enhanced Raman spectroscopy (SERS), coherent Raman scattering (CRS), and multimodal platforms—have transitioned from proof-of-concept demonstrations to impactful tools in biomedical research. These advances span therapeutic monitoring, chemical biology imaging, deep-tissue diagnostics, precision oncology, and multimodal analytics. This article synthesizes the most influential reviews in these areas, highlighting emerging trends, limitations, and future directions.

For Pittcon 2026, the James L. Waters Symposium, scheduled for Monday, March 9, from 2:30 to 4:40 p.m. in Room 221A, turns its focus on Generative artificial intelligence (AI) systems in analytical chemistry, which are increasingly being used for analytical data interpretation, algorithm development, experimental planning, and scientific communication. This article introduces the general concepts of generative AI and its use in spectroscopy.

The 2026 James L. Waters Annual Symposium at Pittcon will focus on the integration of generative AI into analytical chemistry, examining how large language models and AI tools can support method development, data analysis, and chemical measurement while maintaining scientific rigor, validation, and interpretability. Continuing its decades-long tradition of connecting historical perspective with emerging technologies, the symposium will feature presentations from leading chemists and spectroscopists, highlighting both the opportunities and challenges of responsibly incorporating AI into chemical measurement science.

In this article, we illustrate how automated calibration systems and sophisticated algorithms are transforming chemical data into actionable knowledge.

Researchers at the University of Minnesota have developed a dual AI framework that uses synthetic vegetation index data and an advanced Transformer-based segmentation model to significantly improve early detection of winter damage on golf course turf.

Researchers at Washington State University Tri-Cities demonstrate that combining Raman and infrared spectroscopy with convolutional neural networks enables highly accurate, low-cost, and field-ready automated plastic identification.

Artificial intelligence is transforming vibrational spectroscopy by automating calibration, feature extraction, and interpretation across Raman, infrared, near-infrared (NIR), and hyperspectral imaging (HSI) systems. This review of articles highlighted in Spectroscopy during 2025 captures several major developments, spanning data fusion, spectral imaging, and industrial and biomedical applications.

In this Spectroscopy blog post, Yang Liu, a Scientific Liaison for General Chapters at the United States Pharmacopeia (USP), discusses how data-driven methodologies are set to transform spectroscopy in 2026 and beyond, particularly in pharmaceutical analysis.

In 2025, the vibrational-spectroscopy community saw a convergence of deep learning, advanced simulation, and portable instrumentation that materially changed how spectra are interpreted and applied.

A proposed update to United States Pharmacopoeia (USP) <1029> was published in July 2025 for industry comment. What’s changed and are the changes significant?

Top articles published this week include interviews with Paolo Oliveri of the University of Genoa (Italy) and Maryam Shakiba and Santiago Marin of the University of Colorado Boulder, and an inside look at vibrational spectroscopy trends.

A new study from Heilongjiang Bayi Agricultural University introduces a high-accuracy, explainable deep learning model that significantly improves nondestructive nitrogen and chlorophyll estimation in maize canopies using hyperspectral data.