Preprocessing

Preprocessing that happens to added datasets.

General

The OSCA handbook was used to guide preprocessing choices.

Pseudoalignment

💡

kallisto v0.46.0 is used for pseudo-quantification with an index built using GRCh38 release 94.

💡

kallisto bustools is up to 51 times faster than Cell Ranger and runs in constant memory.

Droplet processing

💡

Empty droplets are detected using emptyDrops, which retains distinct cell types that simple knee point thresholds discard. emptyDrops has also been adapted by Cell Ranger V3.

Feature selection

💡

Highly variable genes are within the top 10% of biological variance.

💡

A priori genes of interest can also be specified by subsetting.

Dimensionality reduction and clustering

💡

The top 30 principle components are used to detect clusters and generate UMAP plots.

💡

Clustering uses the leiden algorithm and the resolution parameter can be adjusted.

Marker genes

💡

Wilcoxon rank sum tests calculated with presto are used to sort up-regulated marker genes for each cluster.

Edit this page on GitHub

Preprocessing

General#

Pseudoalignment#

Droplet processing#

Feature selection#

Dimensionality reduction and clustering#

Marker genes#