![](/themes/custom/sib/img/trames/tiny/sm/circle-1-blue.png)
![](/themes/custom/sib/img/trames/tiny/sm/circle-1-green.png)
LS2 Annual Meeting 2024
![](/sites/default/files/styles/header_image/public/2023-10/LS2AM24%20teaser%20text%20and%20QR.png?h=0964da68&itok=MNMyv3FV)
Visit us booth #16!
The LS2 Annual Meeting takes place at the University of Lausanne, 14-15 February 2024. This year's Annual Meeting holds the theme "Exploring the Wonders of Life", reflecting our collective dedication to unraveling the mysteries of the natural world. The three keynote lectures will be given by Prof. Nicole King from the University of California, Berkeley (US), Dr. Samer Hattar from the National Institute of Mental Health (US), and Dr. Pavel Tomancak from the Max Planck Institute of Molecular Cell Biology and Genetics (DE).
SIB holds a symposium "Windows to the Wonders of Life" with keynote speaker David Niklas Wissel (Robinson Lab at UZH and the BoevaLab at ETH).
David Niklas Wissel will present "Machine learning for curator-centric ontology-aware cell-type single-cell annotation":
In recent years, there has been a flurry of work on methods for reference-based single-cell type annotation, especially in the context of emerging cell atlases, such as the Fly Cell Atlas and the Human Lung Cell Atlas. While ample previous research has benchmarked cell-typing methods, it was focused primarily on the academic community, leading us to propose a benchmark tailored to the needs of curators and annotators. We focus on benchmarking datasets annotated with the Cell Ontology (CO), as the most comprehensive and widely used hierarchy for annotating cell types. Our benchmark pays special attention to three aspects especially relevant for curators. First, we consider only methods that can perform hierarchical multi-label classification, i.e., models that will make predictions compatible with CO. Second, our work investigates the ability of different models to leverage CO to better handle rare and novel cell types. Lastly, we perform a user study to understand the impact of calibration and particularly its visualization when “deploying” cell-type annotations to aid curators. We hope that our benchmark can help guide the deployment of more curator-centric cell-type classification pipelines. We open-source our code and newly annotated data, including an easy-to-use snakemake pipeline.
Thursday 15 February 2024, 14:05 - 15:50
Organized by the LS2 Intersection Bioinformatics.
Chaired by Prof. Katja Bärenfaller (University of Zurich / SIB - Swiss Institute of Bioinformatics).
Register by Wednesday 10 January 2024: https://annual-meeting.ls2.ch/2024/registration