The Dutch Chemometrics Society (DCS) is excited to announce the third DCS PhD Spotlight session. There will be three talks in the program, with a diversity of topics within chemometrics and other related disciplines.

The session will be held on December 1st 2022 (Thursday) from 12:00-13:00.


  • mEthAE: an Explainable AutoEncoder for methylation data, by Sonja Katz (Wageningen University).

Despite the wealth of knowledge generated through epigenome-wide association studies (EWAS) and links found with common and rare diseases, analysis of methylomic data remains difficult due its high dimensionality. Therefore, deep learning algorithms, such as autoencoders, are increasingly applied to reduce dimensionality into latent space. Here, we propose a deep unsupervised autoencoder for interpretable dimensionality reduction of methylation data (mEthAE). In a proof of principle experiment, we trained our framework on CpG sites located on chromosome 22 and attained impressive compression of information in only a few latent space features. To enable interpretability, we developed a complementary pipeline capable of identifying CpGs associated with latent features and putting them into a genomic and functional context. This framework can aid in increasing the statistical power of other methods and help interpret their results.

  • Cumulative neutral loss model for fragment deconvolution in high-resolution mass spectrometry data, by Denice van Herwerden (University of Amsterdam).

Non-targeted analysis combined with high-resolution mass spectrometry (HRMS) is the most comprehensive approach to chemically characterize complex samples, including environmental and biological samples. These samples can contain thousands of structurally known and unknown compounds and are often analyzed using liquid chromatography coupled with high-resolution mass spectrometry. For identification of these chemicals, one of the steps is clean-up of the spectral information. For the fragments, this is generally achieved by comparing apex retention times or correlation between peaks. However, neither of these methods utilizes the information from the mass domain. Therefore, a probabilistic cumulative neutral loss model is proposed to deconvolute the fragments based on their mass domain information.

  • A novel approach for analyzing qualitative Chemometrics data: the applicability of Rasch modelling, by Andrea Carnoli (Radboud University)

The analysis of chemical data is intertwined with the research question and is often focused on the quantitative aspect of the data. A common example of chemometrics modeling is the exploratory analysis of a dataset containing the amount of a series of molecules via Principal Component Analysis. Chemical data are often quantitative, but sometimes can be seen as qualitative. For example, Mass Spectrometry data can be binary if we assign “1” each time that a chemical is detected and a “0” otherwise. When the dataset is binary the quantitative approach is not optimal, and the researchers must rely on a qualitative one. In this case, the exploratory analysis of the dataset can be conducted via Non-Linear Principal Component Analysis. However, the results of the method are sample-specific and not always easy to interpret due to the possible multidimensionality of the solution. Here we show a framework, composed by Rasch modeling and Generalized Linear Mixed Effect Models, that can extract group-specific information from a multivariate binary dataset by including the information present in the experimental design. We show that the solution obtained by this framework will provide group-specific information easy to interpret due to its unidimensionality. Furthermore, we show that, through Generalized Linear Mixed Effect Models, it is possible to extend the Rasch model to its multilevel form, allowing to consider random factors possibly present. We anticipate this framework to be a useful tool for chemometricians to add group-specific information to those sample-specific extracted via Non-Linear Principal Component Analysis.


The session will be streamed online via Microsoft Teams, will last around 1 hour, and is free of charge. You do not need to register for this event. The link will be shared the same day of the event via e-mail to all those who are subscribed to the Dutch Chemometrics Society mailing list. You can subscribe to the DCS mailing list here.


More about the DCS PhD Spotlight sessions

As a consequence of the Covid pandemic, most of the scientific conferences and events have followed a short online format with few keynote talks in the program. While this has still allowed professionals and renowned scientific speakers to share their research, PhD students had less opportunities to present their work and get valuable feedback from the scientific community during their first steps in academia.

As it will still take some time to get back to the pre-pandemic situation in international traveling and face to face events, The Dutch Chemometrics Society would like to provide PhD students a platform to share their research around Chemometrics and related disciplines.

For this reason we are excited to announce the new DCS PhD Spotlight sessions, a series of free online seminars where PhD students will present their work.



< Previous page