MLMichael Lau

Michael Lau

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Michael Lau has a total of 14 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Interaction between genetic predisposition to successful ageing and chronic air pollution on lung disease in elderly women: results of the German SALIA cohort

    by Sara Kress, Michael Lau, Claudia Wigmann, Michael J Abramson, Holger Schwender, Tamara Schikowski

    Objective: To investigate the interplay between the genetic predisposition to successful ageing and air pollution on lung disease in healthy aged German women under the hypothesis that ageing and lung diseases share mechanisms of oxidative stress and inflammation that can be regulated by genetic predisposition and environmental factors. Design: German Study on the influence of Air pollution on Lung function, Inflammation and Aging prospective cohort between baseline (1985–1994) and follow-up (2007–2010). Setting: Urban Ruhr area and the adjacent rural Münsterland in Germany. Participants: At baseline, 4874 women aged 55 years living between 1985 and 1994 in the setting and at follow-up examination, 834 of them participated. Main outcome measures: Chronic lung disease was defined as any of asthma, chronic bronchitis, cough (with sputum) or chronic obstructive pulmonary disease. Chronic individual exposures to nitrogen dioxide (NO 2 ), nitrogen oxides, particulate matter with median aerodynamic diameters 2.5 ), PM 10 , PM coarse and PM 2.5 absorbance based on European Study of Cohorts for Air Pollution Effects land-use regression models were used. Main and interaction effects between the genetic risk score (77 single-nucleotide polymorphisms (SNPs) related to successful ageing) and air pollutant exposures were investigated using adjusted logistic regression models. Results: In 560 women (67–80 years), chronic lung disease was present in 156. Higher exposure to air pollution was associated with increased odds by up to 43% per IQR-increase in NO 2 (IQR=11.6 µg/m³, 95% CI 1.15 to 1.77). The genetic make-up reduced the negative impact of air pollution (gene–environment interaction with NO 2 : OR=0.66, 95% CI 0.45 to 0.96), while a healthy lifestyle further strengthens this association. Conclusions: In elderly women, genetic predisposition based on successful ageing SNPs likely reduces the negative impact of air pollution on chronic lung disease, while a healthy lifestyle further strengthens this association.

  • Open Access

    Quantitative and qualitative changes in substance-related administrative offences in road traffic during the SARS-CoV-2 pandemic in Munich

    by Anna Holzer, Andreas Stoever, Michael Lau, Sabine Gleich, Matthias Graw, Anouk Ludwig, Benno Hartung

    Introduction: The SARS-CoV-2 pandemic beginning in 2020 led to significant restrictions on social life and mobility, raising concerns about increased substance use across the general population. To investigate whether the pandemic resulted in quantitative or qualitative changes in alcohol and/or drug use in the context of road traffic, a retrospective analysis of toxicological findings was conducted in the city of Munich, considering the local pandemic-related restrictions. Materials and methods: A total of 6,210 blood samples were analyzed from individuals suspected of committing substance-related administrative traffic offences under §24a of the German Road Traffic Act between January 1, 2019, and July 31, 2021. Samples were examined for the presence of substances, their concentrations, and the type of vehicle involved. The cohort was stratified into pre-pandemic and pandemic periods, with March 16, 2020 set as the cut-off date. The pandemic period was further subdivided based on the severity of imposed restrictions. Statistical comparisons were performed using Fisher’s exact test, t-tests, ANOVA, and logistic regression. Results: Cannabis was the most frequently detected substance (66.2% pre-pandemic; 67.4% during the pandemic), followed by alcohol (11.7% vs. 10.8%) and cocaine (5.7% vs. 5.2%). Only minor differences were observed between the pre-pandemic and pandemic periods, as well as across phases of mild versus severe restrictions. Notably, THC-COOH concentrations were higher during the pandemic. Alcohol levels were elevated during phases of light restrictions and reduced during periods of strict lockdown. Cannabis was most commonly detected in car drivers, whereas alcohol was more frequently found in e-scooter riders, particularly during less restrictive phases. Conclusion: Substance detection patterns among drivers in Munich showed overall stability during the COVID-19 pandemic, with cannabis remaining the most commonly identified drug. However, shifts in substance concentrations and differences by vehicle type and restriction severity suggest subtle changes in consumption behavior. These findings underscore the need for continued surveillance and context-specific traffic safety measures.


  • Open Access

    logicDT: a procedure for identifying response-associated interactions between binary predictors

    by Michael Lau, Tamara Schikowski, Holger Schwender

    Interactions between predictors play an important role in many applications. Popular and successful tree-based supervised learning methods such as random forests or logic regression can incorporate interactions associated with the considered outcome without specifying which variables might interact. Nonetheless, these algorithms suffer from certain drawbacks such as limited interpretability of model predictions and difficulties with negligible marginal effects in the case of random forests or not being able to incorporate interactions with continuous variables, being restricted to additive structures between Boolean terms, and not directly considering conjunctions that reveal the interactions in the case of logic regression. We, therefore, propose a novel method called logic decision trees (logicDT) that is specifically tailored to binary input data and helps to overcome the drawbacks of existing methods. The main idea consists of considering sets of Boolean conjunctions, using these terms as input variables for decision trees, and searching for the best performing model. logicDT is also accompanied by a framework for estimating the importance of identified terms, i.e., input variables and interactions between input variables. This new method is compared to other popular statistical learning algorithms in simulations and real data applications. As these evaluations show, logicDT is able to yield high prediction performances while maintaining interpretability.


  • Open Access

    Efficient gene–environment interaction testing through bootstrap aggregating

    by Michael Lau, Sara Kress, Tamara Schikowski, Holger Schwender

    Gene–environment (GxE) interactions are an important and sophisticated component in the manifestation of complex phenotypes. Simple univariate tests lack statistical power due to the need for multiple testing adjustment and not incorporating potential interplay between several genetic loci. Approaches based on internally constructed genetic risk scores (GRS) require the partitioning of the available sample into training and testing data sets, thus, lowering the effective sample size for testing the GxE interaction itself. To overcome these issues, we propose a statistical test that employs bagging (bootstrap aggregating) in the GRS construction step and utilizes its out-of-bag prediction mechanism. This approach has the key advantage that the full available data set can be used for both constructing the GRS and testing the GxE interaction. To also incorporate interactions between genetic loci, we, furthermore, investigate if using random forests as the GRS construction method in GxE interaction testing further increases the statistical power. In a simulation study, we show that both novel procedures lead to a higher statistical power for detecting GxE interactions, while still controlling the type I error. The random-forests-based test outperforms a bagging-based test that uses the elastic net as its base learner in most scenarios. An application of the testing procedures to a real data set from a German cohort study suggests that there might be a GxE interaction involving exposure to air pollution regarding rheumatoid arthritis.


  • Open Access

    Evaluation of tree-based statistical learning methods for constructing genetic risk scores

    by Michael Lau, Claudia Wigmann, Sara Kress, Tamara Schikowski, Holger Schwender

    Background: Genetic risk scores (GRS) summarize genetic features such as single nucleotide polymorphisms (SNPs) in a single statistic with respect to a given trait. So far, GRS are typically built using generalized linear models or regularized extensions. However, these linear methods are usually not able to incorporate gene-gene interactions or non-linear SNP-response relationships. Tree-based statistical learning methods such as random forests and logic regression may be an alternative to such regularized-regression-based methods and are investigated in this article. Moreover, we consider modifications of random forests and logic regression for the construction of GRS.Results: In an extensive simulation study and an application to a real data set from a German cohort study, we show that both tree-based approaches can outperform elastic net when constructing GRS for binary traits. Especially a modification of logic regression called logic bagging could induce comparatively high predictive power as measured by the area under the curve and the statistical power. Even when considering no epistatic interaction effects but only marginal genetic effects, the regularized regression method lead in most cases to inferior results.Conclusions: When constructing GRS, we recommend taking random forests and logic bagging into account, in particular, if it can be assumed that possibly unknown epistasis between SNPs is present. To develop the best possible prediction models, extensive joint hyperparameter optimizations should be conducted.