Affiliation
Professor in Biostatistics
Dep. Epidemiology and Data Science
Amsterdam University Medical Center, Amsterdam, NL
mark.vdwiel [at] amsterdamumc.nl
www.bigstatistics.nl/mark-van-de-wiel/
LinkedIn
Research statement
Connecting data with mathematics drives most of my statistical research: provide a generic, robust solution for a given study, and one likely solves similar problems for many studies. My research interests cover a wide spectrum, including high-dimensional data analysis (omics) and dedicated shrinkage techniques, but my main fascination nowadays is prediction and classification for medical data by either statistical or machine learners. Here, I focus on developing methods to improve predictive performance and biomarker selection by structural use of complementary data (co-data), e.g. from external studies or data bases. Moreover, we develop tools to aid interpretation of ML, e.g. by providing inference for variable importance metrics, such as Shapley values. We directly apply and test such methods in a number of collaborative projects on cancer diagnostics and prognostics.
Selected presentations
Oslo, September 2024 (XAI workshop): Shapley values and their uncertainties in complex regression models for epidemiological applications
Milan, August 2023 (ISCB): Linked shrinkage to improve estimation of interaction effects in regression models
Cambridge, November 2021 (Armitage workshop): Improving prediction, variable selection and treatment effect estimation by the adaptive, multi-penalty elastic net
Some recent publications (Full list: Google Scholar; Web-of-Science)
van de Wiel, M. A., Amestoy, M., & Hoogland, J. (2024). Linked shrinkage to improve estimation of interaction effects in regression models. Epidemiologic Methods, 13(1), 20230039.
Goedhart, J. M., Klausch, T., & van de Wiel, M. A. (2023). Estimation of predictive performance in high-dimensional data settings using learning curves. Computational Statistics & Data Analysis, 180, 107622.
van Nee, Mirrelijn M., Lodewyk FA Wessels, and Mark A. van de Wiel. "ecpc: an R-package for generic co-data models for high-dimensional prediction." BMC bioinformatics 24.1 (2023): 172.
Andrade Barbosa B, ..., Van de Wiel MA*, Kim, Y* (2021). Bayesian log-normal deconvolution for enhanced in silico microdissection of bulk gene expression data. Nature communications, 12, 1-13.
van Nee, Mirrelijn M, Lodewyk FA Wessels, and Mark A. van de Wiel. Flexible co‐data learning for high‐dimensional prediction. Statistics in medicine 40.26 (2021): 5910-5925.
Rauschenberger, Armin, Enrico Glaab, and Mark A. van de Wiel. Predictive and interpretable models via the stacked elastic net. Bioinformatics 37.14 (2021): 2012-2016.
R Packages
We want our methods to be used, so we implemented these in R-packages, which include data, example(s) and documentation. Group packages.
Teaching
Together with Wessel van Wieringen, I teach High-Dimensional Data Analysis in the Statistical Science Master, Leiden. Topics involve: regularized regression, multiple testing, shrinkage, empirical Bayes, analysis of high-dimensional count data. I also contribute to the Biostatistics course to medicine students in the research master Personalized medicine.
Media
Blogs: VVS-OR website; personal website.
Interview/Comments on Personalized Medicine in the 'Trouw' (Dutch newspaper), 09/12/2017
Nieuwsbericht honorering ZONMW TOP subsidie voor project "Compute CANCER", February 2017.
Contribution 'Nieuw Archief voor de Wiskunde': "Statistiek op het genoom: ‘Big Data’, maar dan anders", December 2015
Interview VOZ magazine, guest-edited by Wouter Bos, July 2015
Interview in the 'Volkskrant' (Dutch newspaper), 23/08/2014
