Spots classification (Machine learning)

José Fernández Navarro edited this page Jan 3, 2018 · 3 revisions

A very useful feature is to classify spots based on their gene expression values. This can be used to measure the quality of the data, to select spots based on cell type or even to redefine anatomical regions.

You can do this by clicking on the "Spots classification" button in the main view. Basically, the spots (all the spots present in the dataset) will be assigned a color (similar color will represent similar expression patterns). The colors are computed by first doing a dimensionality reduction (t-SNE or PCA) and then clustering (KMeans or HClust) the dimensionality reduced coordinates to finally assign one color to each cluster. The number of clusters needs to be defined but you can get an estimation of this using the "Estimate" button.

The Spot classification window allows to choose the dimensionality reduction algorithm and its settings, the normalization method, the clustering method and to define thresholds.

The computation may take few seconds but once it is completed you will see a scatter plot of the clustered spots and you will also see the colored spots on top of the tissue image in the main view.

A very useful feature it so select spots from the scatter plot using the right click of the mouse. This way you can select spots based on the gene expression similarities or simply outliers.

You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.