EuBIC-MS Winter School 2022: Full program

A tour of machine learning

Machine learning has become ubiquitous in mass spectrometry and proteomics (as in many other fields), fueled largely by the increasing availability and amount of data. Learning algorithms can figure out how to perform important tasks by generalizing examples. This workshop is an interactive Jupyter Notebook (Python) that will teach you how to build successful machine learning models. No background in machine learning is assumed, just a keen interest.

Required skills: Models are built in the Python programming language so some basic knowledge of Python is required. This free Python tutorial should give a solid basis to participate in the workshop: https://www.kaggle.com/learn/python.

Monday, full day workshop

Differential abundance analysis for proteomics

Mass spectrometry based proteomics experiments generate ever larger datasets and, as a consequence, complex data interpretation challenges. This hands-on workshop focuses on the statistical concepts for differential analysis of mass spectrometry based proteomics data  acquired via label free data dependent technologies. Moreover, more advanced experimental designs and blocking will also be introduced. The hands-on will rely exclusively on free and user friendly open source tools in R/Bioconductor. The course will provide a solid basis for beginners, but will also bring new perspectives to those already familiar with standard data interpretation procedures in proteomics.

This workshop is a shortened proteomics-oriented version of Lieven’s Practical Statistics for Life Sciences course, taking place at the IGC the week after the Winter School; potentially a good reason to extend your stay in Portugal.

Required skills: Very basic knowledge in statistics and minimal skills in R.

Monday: full day workshop

STRING and Cytoscape for proteomics data analysis

Modern high-throughput technologies, such as proteomics, produce an enormous amount of new data on individual genes and proteins. These, often too long, lists can be interpreted in light of cellular context and existing biological knowledge with the help of protein network resources such as STRING. This workshop will cover the basics of using the STRING database and the network visualization and analysis software tool Cytoscape to analyze such protein lists. Specifically, you will be able to import your proteomics data into Cytoscape using the stringApp, master network layouts and data visualization techniques, and perform clustering and enrichment analyses. 

Required skills: No prior skills are required. To make the best use of the workshop, we encourage you to bring a laptop (not a tablet) with a working web browser and the latest version of Cytoscape (3.9.1) already installed. 

Monday: half day workshop (first half)

Biological interpretation by clustering and protein complex analysis

Standard quantitative proteomics data is often highly dimensional. It comprises multiple sample types with each coming in replicates to account for the biological variation. For such data, the extraction of relevant features for biological interpretation can become sophisticated due to data complexity and difficulties in proper visualisation. This workshop will present methods to find proteins with similar behaviour by clustering their expression profiles, and apply quantitative analysis of protein complexes through their mostly highly co-regulated subunits. You will learn about different clustering methods and their underlying principles. You will furthermore apply them using R scripts and the interactive web services VSClust, ComplexBrowser and CoExpresso.

Required skills: Basic experience with running R scripts is recommended but not necessary. Please bring your laptop with RStudio installed (or Jupyter with an R kernel). Optionally, if you want to run the web services on your own computer, I recommend a Docker environment.

Monday: half day workshop (second half)

Feathering the dinosaurs: Machine learning unveils the proteome-wide modification landscape

As a kid, I was taught to think of dinosaurs as giant lizards, covered in reptilian scales. But more recent discoveries showed that quite a few dinosaurs had feathers, which rather dramatically changed our picture of some of these majestic animals. I remember that being quite a shock, as one’s old beliefs were challenged, and a new version of reality had to be processed and accepted. Interestingly, we may be looking at a similar awakening to new perceptions in proteomics, as we are becoming increasingly aware of the widespread nature of modifications on quite a lot of proteins. It is of particular interest that the key breakthrough to allow this new insight lay not in the wet lab or instrumentation side of proteomics, but purely in the bioinformatics aspect. Indeed, the advent of open modification search engines has begun to reveal a plethora of chemical alterations on proteins, including experiment-induced artefacts, but also biological artefacts as well as meaningful modifications. And it has done so even for existing data!
However, one of the key challenges of such open modification searches revolves around the linked problems of specificity and sensitivity. Our novel ionbot search engine tackles these issues through a multilayer machine learning approach that combines highly accurate predictors of analyte behaviour with fully machine learning scoring based on the target-decoy approach.
With ionbot, we then proceeded to analyse the largest set of public domain proteomics data to date for both human and mouse, revealing for the first time their proteome-wide modification landscape.
Here, ionbot and its components will be briefly discussed, along with a bird’s eye view of the proteome-wide modification landscapes obtained for human and mouse.

Tuesday, 09:10, main auditorium

Proteome wide assessment of PTMs in their structural context

An important mechanism to regulate the activity and function of proteins are post translational modifications (PTMs). Despite impressive technological progress in identifying and quantifying PTMs by mass spectrometry, a persisting challenge is to distinguish functionally relevant modifications from the tens of thousands that are routinely measured. To improve our understanding of functional relevance, we set out to explore the potential of integrating PTM data with proteome wide structural information, that only recently became available based on structure predictions by AlphaFold2. I will discuss the tremendous potential of such integrative analyses based on several examples and I will present a toolset that enables the community to easily analyze and visualize PTMs in their structural context.

Tuesday, 10:15, main auditorium

Direct Comparison of Tandem Mass Spectra: Theory and Applications

A decade ago, Palmblad and Deelder first described a method for molecular phylogenetics based on direct comparison of tandem mass spectra. This method has since seen a range of practical applications, including food and feed species identification, quality control, and optimization of experimental design. Albeit these applications can be considered niches, the method, and compareMS2 software implementing it, nevertheless has a number of unexpected uses.
The compareMS2 software is currently undergoing its first major revision, adding features such as dynamic phylogenetic tree built on-the-fly, a wide range of spectral filters and a perfectly symmetric distance metric, all combined in a simple graphical user interface. The software can also create unique visualizations of MS2 similarity as a function of MS1 (precursor) differences, with possible quality control applications. The software is open source and available on https://github.com/524D/compareMS2.

Tuesday, 11:00, main auditorium

An automated workflow for multiplexed single cell proteomics sample preparation at unprecedented sensitivity

The analysis of single cell proteomes has recently become a viable complement to transcriptomics and genomics studies. Proteins are the main driver of cellular functionality and mRNA levels are often an unreliable proxy of such. Therefore, the global analysis of the proteome is essential to study cellular identities. Both multiplexed and label-free mass spectrometry-based approaches with single cell resolution have lately attributed surprising heterogeneity to believed homogenous cell populations. Even though specialized experimental designs and instrumentation have demonstrated remarkable advances, the efficient sample preparation of single cells still lacks behind. Here, we introduce the proteoCHIP, an universal option for single cell proteomics sample preparation at surprising sensitivity and throughput. The automated processing using a commercial system combining single cell isolation and picoliter dispensing, the cellenONE®, allows to reduce final sample volumes to low nanoliters submerged in a hexadecane layer simultaneously eliminating error prone manual sample handling and overcoming evaporation. The specialized proteoCHIP design allows for the direct injection of single cells via a standard autosampler resulting in around 1,500 protein groups per analytical run at remarkable reporter ion signal to noise while reducing or eliminating the carrier proteome. We identified close to 2,600 proteins across 170 multiplexed single cells from two highly similar human cell types. This dedicated loss-less workflow allows to distinguish in vitro co-differentiated cell types of self-organizing cardiac organoids based on indicative markers across 150 single cells. In-depth characterization revealed enhanced cellular motility of endothelial cells and acute myocardium sarcomere organization in cardiomyocytes. Our versatile, and automated sample preparation has not only proven to be easily adoptable but is also sufficiently sensitive to drive biological applications of single cell proteomics.

Wednesday, 09:10, main auditorium

Developing and maintaining tools in a multidisciplinary research team

After 5 years in a bioinformatics research group, moving to an environment dominated by wet-lab scientists required some adaptation. In this talk, I will not only present some new tools developed at the Kusterlab for improving peptide and protein identification, but also discuss some of the challenges we faced in bringing these tools to the wet-lab scientists. The first tool, SIMSI-Transfer, uses MS2 spectrum clustering to reduce missing values across TMT-batches. This is especially relevant for the personalized treatment recommendation studies we run in collaboration with clinics around Germany. Using this tool, we identified up to 40% more PSMs and 9% more proteins per TMT-batch in a large CPTAC cohort study. The second tool, Picked Protein Group FDR, generates accurate and sensitive FDR estimates on the protein group level and scales well up to the repository scale. It allows easy combination and comparison of multiple large-scale studies without the need to re-search them in one analysis. Applying this tool to the entire human section of ProteomicsDB resulted in evidence for over 1,300 genes to have multiple protein isoforms. Finally, I will present some of the solutions we implemented to provide our lab members easy access to these tools, as well as to other tools developed in our lab, such as Prosit and ProteomicsDB.

Wednesday, 10:15, main auditorium

SIRIUS and beyond: Turning tandem mass spectra into metabolite structure information

Liquid Chromatography Mass Spectrometry is a highly sensitive experimental platform for the analysis of metabolites and other small molecules. Unfortunately, structural elucidation of metabolites from tandem MS data remains highly challenging; in untargeted metabolomics experiments, only a small percentage of spectra can be annotated via spectral libraries. For more than a decade, my group has been developing computational solutions for this task. In my talk, I will focus on three aspects of our work: Firstly, how to annotate metabolite MS/MS spectra using fragmentation trees via SIRIUS; second, how to search a MS/MS spectrum in a molecular structure database via CSI:FingerID; and third, how to assign thousands of compound classes to each spectrum even when the compound is missing from all structural and spectral databases (CANOPUS). I will also quickly touch upon the annotation of isotope patterns, and assigning confidence search results in a molecular structure database via COSMIC.

Wednesday, 11:00, main auditorium

The function of protein post-translational modifications

Hundreds of thousands of protein modifications have been identified to date in human proteins with most not having a characterized function. Given this gap there is a pressing need to develop approaches to study the functional relevance of post-translational modifications. I will introduce some of the recent ideas and methods developed in different labs to tackle this challenge. These include conservation based methods, machine learning approaches and experimental methods including mass spectrometry based ones. After this introduction we will discuss some of the open questions and other potential ways to address this problem.

Wednesday, 13:00, main auditorium

Lennart Martens

Ghent University & VIB, BE

Lennart Martens is Full Professor of Systems Biology at Ghent University, Group Leader of the Computational Omics and Systems Biology (CompOmics) group at VIB, and Associate Director of the VIB-UGent Center for Medical Biotechnology, all in Ghent, Belgium. He has been working in proteomics bioinformatics since his Master’s degree, which focused on the computational interpretation of peptide mass spectra, and the sequence-dependent fragmentation of peptides. He then worked as a software developer and framework architect for a software company for a few years, before returning to Ghent University to pursue a Ph.D. in proteomics and proteomics informatics. During this time, he worked on the development of high-throughput peptide centric proteomics techniques and on bioinformatics tools to support these new approaches. In 2003 he designed and built the PRIDE repository for the global dissemination of proteomics data at EMBL-EBI as a Marie Curie fellow of the European Commission. After obtaining his Ph.D., he rejoined EMBL-EBI to coordinate the newly created PRIDE group for the next three years, firmly establishing the system as the world’s foremost public proteomics data repository. He then moved back to Ghent University and VIB to take up his current positions, in which he focuses on novel machine learning algorithms for mass spectrometry data analysis, and their application to the large-scale orthogonal reprocessing of public proteomics data. Prof. Martens has been elected President of the European Proteomics Association (EuPA) in 2020.

Isabell Bludau

Max Planck Institute of Biochemistry, DE

Isabell Bludau is a postdoctoral fellow in the lab of Prof. Matthias Mann at the Max Planck Institute of Biochemistry near Munich. During her PhD with Prof. Ruedi Aebersold at the Institute of Molecular Systems Biology, ETH Zurich, Isabell developed computational methods for analyzing large-scale proteomics data. She specifically focused on the detection and quantification of protein complexes. Recently, Isabell’s work focuses on the inference of different proteoforms and their crosstalk. She is also an active developer of different tools within the AlphaPept ecosystem. Isabell’s PhD thesis was awarded with the ETH silver medal and her postdoctoral research is supported by a Postdoc.Mobility fellowship of the Swiss National Science Foundation.

Magnus Palmblad

Leiden University Medical Center, NL

Magnus Palmblad is Associate Professor and Head of Bioinformatics at the Center for Proteomics and Metabolomics of the Leiden University Medical Center (LUMC). He did his MSc in Molecular Biotechnology and PhD in Ion Physics at Uppsala University in 2002. After a 3-year postdoc at the Center for Accelerator Mass Spectrometry at Lawrence Livermore National Laboratory, he was appointed as Senior Research Fellow at the University of Reading’s BioCentre and joined the LUMC Biomolecular Mass Spectrometry Unit in 2007. His main research interests are in computational proteomics, including novel algorithms and reproducible data analysis workflows. Other research lines are applications of direct comparisons of tandem mass spectra and systematic analysis of the biomedical literature combining text mining, bioinformatics and machine learning. Dr. Palmblad has served on Editorial Boards of several journals, including the Journal of Proteome Research.

Karl Mechtler

IMP, Vienna Biocenter, AT

Karl Mechtler serves as Head of the Protein Chemistry Facility at the IMP, IMBA, and GMI in Vienna, Austria, since 2000 and is Head of Mass Spectrometry at VBCF since 2010. He received the Lower Austrian Science Award in 2004 and the Outstanding Scientist Technology Award from ABRF in 2010. 2021 Karl was honored with the APMA Lifetime Award. From 2007 to 2012 Karl Mechtler was leading member of the Proteomics Standards Initiative of the Association of Biomolecular Resource Facilities and from 2011 to 2015 he was elected President of the Austrian Proteomics Society (APMA). Karl was very active in advancing proteomics education as he served as coordinator of the Education Committee of the European Proteomics Association (EuPA). Together with Lennart Martens Karl also organized the first Proteomics Bioinformatics Conference in Europe (Semmering, Austria), which was the starting point of all EuBIC Winter Schools.

Matthew The

Technical University Munich, DE

Fascinated by mathematics and applying it to complex problems that do some good for the world, Matthew The has found his challenge in applying statistical and machine learning methods to biological and clinical data. He obtained a PhD in the lab of Lukas Käll in Stockholm, maintaining the software package Percolator and developing clustering algorithms for mass spectrometry data (MaRaCluster, Quandenser). Now, he is heading the bioinformatics team in the lab of Bernhard Küster in Munich, where they develop tools to visualize and extract relevant information from the huge amounts of MS data generated every day (ProteomicsDB, Prosit). In his spare time, Matthew enjoys learning new languages, playing chess, traveling and last but not least, cooking and eating delicious food.

Sebastian Böcker

FSU Jena, DE

Sebastian Böcker holds the Chair for Bioinformatics at the Institute for Computer Science, Friedrich Schiller University Jena, Germany. He studied mathematics and did his PhD in biomathematics at Bielefeld University, focusing on theoretical phylogenetics. He then went to industry for three years, developing computational methods for the interpretation of DNA/RNA mass spectrometry data. He returned to Bielefeld University as an independent research leader, before he took up his current position in Jena. His research interests are mainly method-driven and were originally focussed on combinatorics and algorithmics; later, stochastics and machine learning joined the methods of interest. On the application side, his research focuses on the annotation of small molecules from mass spectrometry data; in 2020, SIRIUS 4 from his group was named “method to watch” by Nature Methods. Sebastian Böcker is a Emmy Noether fellow (Computer Science Action Program) of the Deutsche Forschungsgemeinschaft and also a fellow of the Alexander-von-Humboldt Society. In his spare time, Sebastian likes to sit on his couch and binge watch.

Pedro Beltrao

ETH Zürich, CH

Since January 2022, Pedro Beltrao is Professor for Computational Biology at ETH Zürich. Before that, he was group leader at EMBL-EBI within the Wellcome trust Genome Campus in Cambridge, UK. Pedro and his team worked on the consequences of genetic variation on evolution & disease. He is particularly well known for pushing our understanding on the function of post translational modifications within regulation networks.

Aurelien Dugourd

Heidelberg University, DE

Aurélien Dugourd graduated in 2013 with a first master’s degree in Genetics, Genomics and Biotechnology at the University de Bretagne Occidentale. He then pursued a second master’s degree in computational biology at the University of Nantes. After his second graduation in 2015, he joined Julio Saez-Rodriguez’s team at the JRC COMBINE lab in Aachen as a PhD student. He worked on the development of a hybrid mechanistic model, integrating gene regulation, signaling pathways and metabolomics data to explain disease phenotypes, help find new therapeutic targets and predict their potential effect based on a specific patient profile. This project was part of the collaborative SyMBioSys ITN project, financed by the European Marie Sklodowska-Curie actions. Currently, Aurélien Dugourd is a postdoc in Julio Saez-Rodriguez’s lab working primarily in the SMART-CARE consortium to apply mass spectrometry-based systems medicine to cancer.

Viktoria Dorfer

Fachhochschule Oberösterreich, AT

Viktoria Dorfer studied Bioinformatics at the University of Applied Sciences Upper Austria and received her Ph.D. in informatics from the Johannes Kepler University Linz. Her research interests focus on computational proteomics, especially on peptide identification, which was also the topic of her Ph.D. thesis, entitled “Identification of Peptides and Proteins in High-resolution Tandem Mass Spectrometry Data”. Part of this thesis was the development of the MS Amanda peptide identification algorithm. At present, Vikoria is working as Professor for Bioinformatics at the University of Applied Sciences Upper Austria and is supervising several Ph.D. students and master students in the field of computational proteomics. In addition, she is a board member of the Austrian Proteomics and Metabolomics Association (APMA) and an active EuBIC-MS member.

Marc Vaudel

University of Bergen, NO

Marc Vaudel is a research group leader at the Mohn Research Center for Diabetes Research, Haukeland University Hospital and University of Bergen, Norway, and senior researcher at the Norwegian Institute of Public Health. Can some children with diabetes avoid insulin shots? How does pregnancy and life as a baby influence the risk of later obesity and diabetes? Marc’s research group answers these questions through the study of the interplay between genetics, biological signaling, and environmental factors. Marc has been involved in the EuBIC activities since the very early days, and a firm believer in community-driven scientific initiatives like EuBIC.