 Welcome to this short introduction to the core concepts of single cell and spatial omics. Omics is fundamentally about measuring all the things in a sample. I've earlier talked about transcriptomics, which measures all the transcripts in a sample, and MS-based proteomics, which measures all the proteins in a sample using mass spectrometry. When it comes to single cell omics, this only differs in the sense that we assign things to cells, meaning that we know which things come from the same cell, and therefore can make an inventory of each cell. Spatial omics, similarly, is about assigning coordinates for things, so that we know which things were close to each other in the sample. Let's start with the typical single cell workflow. The starting point is usually a tissue sample, and the first step is to create a so-called dissociated cell culture in which the cells don't stick to each other. That's important for the next step, which is to isolate single cells. This can be done using either cell sorting with techniques such as fax or max, laser capture microdice action, or microfluidics, which allow you to create micro droplets with cells in separate droplets. The latter is done with a setup like this. Imagine you have water coming in with cells in it, and you have oil coming in the other direction, pinching off tiny droplets that each contain at most one cell. Once you've isolated the cells, the obvious thing to do is simply to do omics on each cell, and if you have sufficiently sensitive methods, that is possible. The problem is it doesn't scale very well, so this approach scales to maybe some hundreds of cells, but not much beyond that. The alternative is to label the cells, that is put so-called unique molecular identifiers on all the things from a cell. A good example of that is DNA barcodes where you label, for example, all the transcripts in a cell with a DNA barcode, and the transcripts from another cell with a different DNA barcode. Once you've labeled all the things, you can mix the cells and analyze all of it in multiplexed omics, doing, for example, next-generation sequencing. The trick is that this scales to tens of thousands of cells, and it has the further advantage that since all of them are analyzed in parallel in one batch, you avoid the usual issues of batch effects where cells may not be directly comparable. So how about the spatial workflow? If we're looking at spatial transcriptomics, the typical approach is in situ capture. What you do is you make an array of beats which have been DNA barcoded. You put the beats down randomly, and then do spatial barcode indexing to find out which barcodes correspond to which positions in the array. The next step is to take a tissue section, so a very thin tissue slice, and transfer the RNA to the beats, and then in a clever trick produce cDNA from the transcripts that become barcoded with the barcode from the beat that the RNA was stuck to. That means that you now know for each cDNA which beat it came from and thereby the position. You can now again simply do multiplex sequencing, and you get transcripts with associated positions. The problem is that you can have multiple cells per beat so you don't necessarily know which transcripts are from the same cell. However, recently it has become possible to make small enough beats that statistically speaking you have single cell resolution, and that way you produce single cell spatialomics data. An alternative approach is laser capture micro dissection. You may not think about it, but when you're cutting out the cells using a laser, you inherently know the location of the cell, and for that reason it is spatialomics. So now that we have the data, how do we do the computational analysis of it? When it comes to analyzing single cell data versus bulk data, the most common complaint I hear is poor coverage or that the data are very sparse. And that's because in each cell you may not observe many, for example, transcripts, and when you look at a transcript in most cells it will not have been seen. So you have a matrix with many zeros. But this is bit of an unfair criticism in my opinion because it conflates cells with samples. Cells are not samples when it comes to single cell data, because from one sample you will get many cells. And if I wanted to, I could pool all the cells from a sample and I would get bulk data with excellent coverage. The problem with that approach is obviously that I would completely ignore the cell information, thereby making it pointless to do single cell data in the first place. For that reason what people typically do is to instead assign cell types using clustering algorithms to group together cells that are similar and laying them out doing Disney plots or UMAP plots to identify how many different cell types you have in the sample and assign cells to these cell types. This then allows you to pool not all the cells, but just the cells of the same type and that way obtain cell type specific data from the single cell data. Another thing that people often want to do with single cell data is trajectory inference, also known as pseudo-time reconstruction. The idea is to identify an ordering of cells so that you get as smooth a transition as possible from one cell to the next in expression space. That way you get a kind of fake time axis showing how cells could be converted into other cells that you've seen in the same sample, thus effectively getting a pseudo-time series. Although it is of course not time series data. When it comes to analyzing spatialomics data, what we're dealing with is data with 2D coordinates assigned to each thing. For this reason it shouldn't come as a surprise that most methods are based on image analysis techniques which allow you to identify regions of similar cells and that way patterning within tissues which can be very interesting especially in developmental biology. That's all I have to say about single cell and spatialomics data. If you want to learn more aboutomics data, I suggest you have a look at this presentation next. Thanks for your attention.