---
id: "claim:modular-efficient-and-constant-memory-single-cell-rna-seq-preprocessing-2021-04-60a02544e2:f24c381425cfc412f480"
kind: "claim"
title: "Empty droplets were removed from the kallisto-bustools E18 mouse count matrix with DropletUtils."
description: ""
url: "https://sina.bio/claim/empty-droplets-were-removed-from-the-kallisto-bustools-e18-mouse-count-27f064d7"
---

# Empty droplets were removed from the kallisto-bustools E18 mouse count matrix with DropletUtils.

## Content

Empty droplets were removed from the kallisto-bustools E18 mouse count matrix with DropletUtils.

## Metadata

```json
{
  "claim_position": 102,
  "source_publication_id": "version:modular-efficient-and-constant-memory-single-cell-rna-seq-preprocessing--2021-04-01-nature-biotechnology",
  "source_work_id": "work:modular-efficient-and-constant-memory-single-cell-rna-seq-preprocessing"
}
```

## Connections (1)

### Modular, efficient and constant-memory single-cell RNA-seq preprocessing

- ID: `version:modular-efficient-and-constant-memory-single-cell-rna-seq-preprocessing--2021-04-01-nature-biotechnology`
- Kind: `publication`
- Direction: `incoming`
- URL: <https://sina.bio/publication/modular-efficient-and-constant-memory-single-cell-rna-seq-preprocessing>

This publication contains this grounded claim.

> Comparative analysis of the 10x Genomics E18 mouse dataset. Analysis of the Cell Ranger and kallisto bustools preprocessed datasets was performed in R. The DropletUtils package33,34 was used to remove empty droplets from the kallisto bustools gene count matrix. For Cell Ranger, the filtered matrix was used. After filtering, genes not detected in any remaining Cell Ranger or kallisto bustools barcode were removed. Seurat was used for basic analysis. First, data were normalized by dividing the UMI count of each gene in each cell by the total UMI counts of that cell, and the result was multiplied by 10,000. Then a pseudocount of 1 was added, and the natural log transform was applied. Subsequently, the normalized data were scaled so the distribution of the expression of each gene would have a mean of 0 and a standard deviation of 1. Subsequently, 3,000 genes with highly variable expression levels were selected with the vst method in Seurat. Then principal component analysis was performed on these genes in the scaled data with the R package irlba called by Seurat. The first 40 principal components were used for t-SNE, which was performed with the R package Rtsne called by Seurat. Clustering was performed with the Leiden algorithm29 on the kallisto bustools and Cell Ranger matrices. The clustering parameters were 20 nearest neighbors and resolution 1. Differential expression analysis was performed with the logistic regression method described in Ntranos et al.31 as implemented in Seurat and applied to the normalized (unscaled) data. Spearman and Pearson correlations were computed for the top 15 cluster marker genes. GSEA was performed on cluster marker genes with adjusted P <0.05 using the R package topGO35 with the GO36,37 annotations provided by Bioconductor 3.10. SingleR38 was used to annotate cell types based on correlation profiles with bulk RNA-seq from ref.39 . Then, the neuronal cell types were used for pseudotime analysis. Pseudotime analysis was performed with slingshot40 via the Docker container from dyno41 .

Source: 2021-04-01-nature-biotechnology.pdf — <https://sina.bio/source/2021-04-01-nature-biotechnology-pdf>