Logo Image
Blog|Podcasts|May 15, 2026

Ep. 45: Overfitting in Chemometrics: Designing Models That Truly Work

Welcome to “Analytically Speaking,” the podcast from LCGC International and Spectroscopy.

Here in Episode #45, podcast host Dr. Jerome Workman speaks with Dr. Rasmus Bro, who is a full professor at the University of Copenhagen and one of the foremost living experts in data analytics and chemometrics, with one of the highest number of reference citations in the field.

We spoke to Prof. Bro about the world of data analysis used for spectroscopy and other analytical methods, with a particular focus on the critical issue of overfitting in calibration modeling. Over the years, he has worked on many aspects of chemometrics, developing numerous algorithms and methods ranging from fuzzy logic and deep learning to analysis of variance and tensor modeling. He has received multiple awards in chemometrics and the analytical sciences, and is the second most cited scientist in the field, with nearly 49,308 citations and an h-index of 86 (Google Scholar). Most of the algorithms and datasets he has developed have been made publicly available to the scientific community.

In this episode, we discuss his 2025 peer-reviewed paper, How to Overfit, published in Chemometrics and Intelligent Laboratory Systems, which examines the persistent problem of overfitting in modern data science, particularly within chemometrics. Prof. Bro explains how the increasing accessibility of machine learning tools has led to widespread misuse and misleading models, and outlines a “caricature protocol” for generating poor models. He highlights key pitfalls such as excessive and poorly understood variables, low-quality or semi-quantitative data, small sample sizes, correlated features, overly complex modeling approaches, and flawed validation practices. He emphasizes that many seemingly predictive models are artifacts of noise, indirect correlations, or poor experimental design rather than true analytical relationships, and that validation alone cannot compensate for inappropriate model complexity or improper use of test sets and cross-validation.

Ultimately, Prof. Bro advocates for a proactive philosophy of “validity by design,” where strong experimental planning, domain knowledge, appropriate sampling, and rigorous validation are built into the modeling process from the outset to ensure robust, interpretable, and scientifically sound results.

References and Further Reading

Weblinks for Rasmus Bro’s work:

(1) Rasmus Bro How to Overfit. Chemom. Intell. Lab. Syst. 2025, 264, 105461. https://doi.org/10.1016/j.chemolab.2025.105461.

(2) Freja Hjertholm; R. Goetz; P. A. A. Schneider; M. A. Petersen; Rasmus Bro; B. Quintanilla-Casas Uncorking White Wine Liking: Combining Analytical Chemistry and Chemometrics with Crowd-Sourced Data to Predict Quality Ratings. Food Chem. 2025, 492, 145376. https://doi.org/10.1016/j.foodchem.2025.145376.

(3) Rasmus Bro; Age K. Smilde Principal Component Analysis. Anal. Methods 2014, 6 (9), 2812–2831. https://doi.org/10.1039/C3AY41907J.

(4) Age K. Smilde; Rasmus Bro; Paul Geladi Multi-Way Analysis: Applications in the Chemical Sciences; John Wiley & Sons: Chichester, U.K., 2005.

(5) Bro, R. PARAFAC. Tutorial and applications. Chemom. Intell. Lab. Syst. 1997, 38 (2), 149–171. https://doi.org/10.1016/S0169-7439(97)00032-4

(6) Rasmussen, M.A.; Rinnan, Å.; Risum, A.B.; Bro, R. Who is winning? A comparison of humans versus computers for calibration model building. J. Chemom. 2021, 35 (12), e3378. https://doi.org/10.1002/cem.3378

(8) Halberg, H.F.F.; Holst, A.Y.; Kaufmann, N.; Bro, R. Calibration model fusion. J. Chemom. 2021, e3350. https://doi.org/10.1002/cem.3350

(9) Bro, R.; Vidal, M. EEMizer: Automated modeling of fluorescence EEM data. Chemom. Intell. Lab. Syst. 2011, 106 (1), 86–92. https://doi.org/10.1016/j.chemolab.2010.06.005

(10) Johnsen, L.G.; Skou, P.B.; Khakimov, B.; Bro, R. Gas chromatography–mass spectrometry data processing made easy. J. Chromatogr. A 2017, 1503, 57–64. https://doi.org/10.1016/j.chroma.2017.04.052

More about our hosts:

Dwight Stoll, PhD:

Dwight R. Stoll is a professor of chemistry at Gustavus Adolphus College in St. Peter, Minnesota. He received his PhD from the University of Minnesota, under Professor Peter Carr, working on the development of fast, comprehensive two-dimensional liquid chromatography (2D-LC). Stoll’s current primary research focus is on the development of 2D-LC for both targeted and untargeted analyses. Active research projects in his laboratory touch on most aspects of multidimensional separation methodologies, including optimization strategies, characterization of selectivity in reversed-phase LC, instrument development, and applications in biopharmaceutical analysis. Stoll is the author or co-author of more than 80 peer-reviewed publications and six book chapters and has instructed numerous short courses in 2D-LC. In 2011 he was the recipient of LCGC’s Emerging Leader in Chromatography Award. In 2017 he received the Georges Guiochon Faculty Fellowship, and was recognized with an Agilent Technologies Thought Leader Award. He is also a member of LCGC’s editorial advisory board and is the editor of the “LC Troubleshooting” column in LCGC.

Jerome Workman, Jr., PhD:

Jerome (Jerry) J. Workman, Jr. is the Executive Editor for LCGC and Spectroscopy. He has held positions as CTO, executive VP, senior research fellow, director, and senior scientist at companies of all sizes, from start-ups to world-leading corporations. He has been an adjunct faculty member of four universities and advised multiple graduate students. He has more than 75 U.S. and international patent applications and 30 issued U.S. and international patents and multiple trade secrets, as well as 500+ technical publications, and 20 reference book volumes on a broad range of spectroscopy and data processing techniques. He has received multiple awards from scientific societies, and has taught annual courses in spectroscopy, chemometrics, and statistics for the AOAC, ACS, ISA, FACSS, and at several universities and corporations. He is a Fellow of the American Institute of Chemists (FAIC), the American Society for Testing and Materials (ASTM), and the Royal Society of Chemistry in the UK (FRSC, CChem, CSci). Jerry holds B.A and M.A degrees from Saint Mary's University of Minnesota, and a PhD degree from Columbia Pacific University working in near-infrared spectroscopy. He is an alumnus of both Columbia University Business School and the MIT Sloan School of Management.

About the Analytically Speaking Podcast:

Analytically Speaking, the podcast from LCGC and Spectroscopy, addresses important issues in separation science and analytical spectroscopy. Topics include new analytical techniques, methods, and approaches; the latest trends; advances in instrument and software technology; practical solutions for specific applications; recent papers in the scientific literature and their applicability; challenges and solutions for data analysis and interpretation; analytical chemistry theory and fundamentals (from advanced research to tutorials and troubleshooting); and more. Our regular hosts are Dwight Stoll, PhD, a professor of chemistry at Gustavus Adolphus College in St. Peter, Minnesota, and Jerry Workman, PhD, a spectroscopist, noted author, and currently the Senior Technical Editor of Spectroscopy and LCGC. Dwight covers separation science and Jerry addresses spectroscopy related topics.

Connect with us on social media:

LCGC LinkedIn page

LCGC on Twitter

LCGC on Facebook

Spectroscopy LinkedIn page

Spectroscopy on Twitter

Spectroscopy on Facebook

Subscribe to LCGC or Spectroscopy magazine or our newsletters.

Email us at [email protected].

To advertise on the show, contact Ed Fantuzzi at [email protected].

Related Content