Introduction to Processing and Analysis of Spatial Multiplexed Proteomics Data

Session 3 – 02:30:00 – Segmentation algorithms to extract single-cell information.
Cell segmentation has made incredible progress thanks to the implementation of machine learning and neural networks in solving this task. There is a zoo of segmentation algorithms out there, but what is the best for your data? We discuss nuances between algorithms to help you decide what might work better for your data and QC techniques.

Break – 00:30:00

Session 4 – 02:30:00 – Practical example of cell segmentation algorithms and QC.
In this session, we navigate through the zoo of popular cell segmentation algorithms, including Cellpose, Mesmer, and Stardist. We will try practical implementations of these algorithms and perform QC checks on the obtained masks.

Session 5 – Duration: 02:30:00 – Image visualization, image formats, and multi-resolution images.
Images are made to be seen. Multiplexed imaging produces tens to hundreds of gigabytes of visual information; therefore, it is essential to understand which tools and structures your data should have to visualize it effectively. In this session, we explain what pyramidal images are, how image visualizers like Napari enable smooth navigation through your image, and how to explore your image to annotate cell types.

Break – 00:30:00

Session 6 – Duration: 2:30:00 – Cell phenotyping
Assigning a cell state or phenotype to the cells in your image is a crucial part of analysing your multiplex proteomics dataset. It facilitates the interpretation of the results in downstream analysis. We will use Napari and the Python library scimap to perform cell-type annotations via intensity gating. In a nutshell, this means defining whether a cell expresses or not particular marker(s) by setting a threshold on the mean intensity of that marker.

Additional tools such as cross_correlation() and freq_DTW() will be used to compare signals through cross-correlation and dynamic time warping techniques. Participants will also learn to calculate signal-to-noise ratios with sig2noise() and identify pitch inflections using inflections(). The session concludes with song_analysis() to examine acoustic patterns at higher hierarchical levels, such as entire songs or vocal sequences.

Session 7 – Duration: 02:30:00 – Introduction to cell neighbourhood analysis I
Spatial omics technologies capture the spatial interactions of the cells in their native context. In this session, we will explore the available methods and their applications to specific questions. We will explore the diversity of different spatial-driven analyses and focus on neighborhood analysis techniques.

Break – 00:30:00

Session 8 – Duration: 02:30:00 – Introduction to cell neighbourhood analysis II
In this practical session, we will perform neighbourhood analysis using methods that can capture the global and local relationships of the tissue. Specifically, COZI is a technique we’ll use to infer statistically enriched neighbour preferences. And Kasumi, a method for identifying spatially localized neighborhoods of intra- and intercellular relationships that are consistent across samples and conditions.

Session 9 – Duration: 02:30:00 – Wrap-up and course summary
In this session, we will discuss which pipelines are currently available for spatial omics analysis. We will also discuss the data frameworks used in the community, as well as provide miscellaneous tips and tricks.

Break – 00:30:00

Session 10 – Duration: 02:30:00 – Wrap-up and questions
Repetition is key to learning. In the last two hours of our course, we will briefly revisit key concepts, clarify general questions from participants, and answer questions related to their specific use cases.