velocity.md

RNA Velocity analysis is a trajectory analysis based on spliced/unspliced RNA ratio.

It is quite popular https://www.nature.com/articles/s41586-018-0414-6,
however, the original pipeline is not well supported: https://github.com/velocyto-team/velocyto.R/issues

There is a new one from kallisto team: https://bustools.github.io/BUS_notebooks_R/velocity.html

1. Install R4.0 (development version) on O2

• module load gcc/6.2.0
• installed R-devel: https://www.r-bloggers.com/r-devel-in-parallel-to-regular-r-installation/
  because one of the packages wanted R4.0
• configure R with ./configure --enable-R-shlib for rstudio
• remove conda from PATH to avoid using its libcurl
• module load boost/1.62.0
• module load hdf5/1.10.1
• installing velocyto.R: velocyto-team/velocyto.R#86

2. Install velocyto.R with R3.6.3 (Fedora 30 example)

bash:

sudo dnf update R
sudo dnf install boost boost-devel hdf5 hdf5-devel
git clone https://github.com/velocyto-team/velocyto.R

rstudio/R:

BiocManager::install("pcaMethods")
setwd("/where/you/cloned/velocyto.R")
devtools::install_local("velocyto.R")

3. Generate reference files

• Rscriptdev 01_get_velocity_files.R
• output:

cDNA_introns.fa
cDNA_tx_to_capture.txt
introns_tx_to_capture.txt
tr2g.tsv

4. Index reference

This step takes ~1-2h and 100G or RAM:
sbatch 02_kallisto_index.sh

• inDrops3 support: BUStools/bustools#4

5. Split reads by sample with barcode_splitter

• merge reads from multiple flowcells first
• https://pypi.org/project/barcode-splitter/

barcode_splitter --bcfile samples.tsv Undetermined_S0_L001_R1.fastq Undetermined_S0_L001_R2.fastq Undetermined_S0_L001_R3.fastq Undetermined_S0_L001_R4.fastq --idxread 3 --suffix .fq

kallisto bus counting procedure works on per sample basis, so we need to split samples to separate fastq files, and merge samples across lanes.

• split_barcodes.sh

6. Count spliced and unspliced transcripts

• kallisto_count
• output:

spliced.barcodes.txt
spliced.genes.txt
spliced.mtx
unspliced.barcodes.txt
unspliced.genes.txt
unspliced.mtx

7. Create Seurat objects for every sample

• create_seurat_sample.Rmd
• also removes empty droplets

8. Merge seurat objects

• merge_seurats

9. Velocity analysis

• velocity_analysis

10. Plot velocity picture

• plot_velocity

11. Repeat marker analysis

• velocity_markers

11. References

• https://www.kallistobus.tools/tutorials
• https://github.com/satijalab/seurat-wrappers/blob/master/docs/velocity.md
• preprocessing influences velociy analysis

Velocity analysis in Python:

• http://velocyto.org/
• https://github.com/pachterlab/MBGBLHGP_2019/blob/master/Figure_3_Supplementary_Figure_8/supp_figure_9.ipynb