Which pipeline to choose for RNA velocity

Question

[Meghamsh]

Dear all, I'm trying to run RNA velocity on my data using ScVelo. I have used velocyto to generate loom files and the scRNA seq analysis was done using Seurat. There are various methods I can proceed to run velocity on my data.

Briefly,

• Merge data -> extract umap embedding for individual condition-> Run ScVelo using individual loom files
• Merge data -> extract umap embedding for all conditions -> Run ScVelo using individual loom files
• Merge data -> extract umap embedding for all conditions -> Run ScVelo using merged loom files
• Merge data -> extract umap embedding for individual condition-> Run ScVelo using merged loom files
• All the above variations with integration of data instead of merging.

I'm not sure if there are multiple ways to arrive at the same interpretation. Which one of these pipelines are right and which ones are absolutely wrong.

Thanks!

Answer 1

[rpolicastro]

RNA-velocity is very sensitive to the manifold it's being run on, and often times cells will separate on the manifold based not on biological variation but on batch effect. Because of this I would integrate the data first to reduce the batch effect.

[Meghamsh]

Thanks for the input. Considering I still go with integrated data, which of the pipeline would you recommend ? I have 3 conditions in my data.

I can integrate them and extract the embeddings. Combine loom files and run velocity

Or

I can integrate, extract umap of individual conditions, run velocity on each of them using individual loom files.

Which one to choose and why ?

[crazyscientist]

Hello rpolicastro and @Meghamash, Thank you for discussing this problem. I would like to get a guidance from you in case of analyzing RNA velocity of scRNA seq samples. I can explain the process in following way-

1. I analzyed scRNA seq data by using Seurat package. I made seuratobject by integrating 3 samples replicates (groupA).
2. Then, I made UMAP, identified cell types for cluster and RNA counts of seuratobject.

3. Next, I followed the following process to convert my suratobject into loom file Converting to/from loom The loom format is a file structure imposed on HDF5 files designed by Sten Linnarsson’s group. It is designed to efficiently hold large single-cell genomics datasets. The ability to save Seurat objects as loom files is implemented in SeuratDisk For more details about the loom format, please see the loom file format specification. pbmc.loom <- as.loom(pbmc, filename = "../output/pbmc3k.loom", verbose = FALSE) pbmc.loom

4. After making loom file, I used the following protocol to get RNA velocity. http://htmlpreview.github.io/?https://github.com/satijalab/seurat-wrappers/blob/master/docs/scvelo.html

However I got an error - reading loom file via hdf5r... Error in H5File.open(filename, mode, file_create_pl, file_access_pl) : HDF5-API Errors: error #000: ../../../src/H5F.c in H5Fopen(): line 429: unable to open file class: HDF5 major: File accessibility minor: Unable to open file

error #001: ../../../src/H5Fint.c in H5F_open(): line 1803: unable to read superblock
    class: HDF5
    major: File accessibility
    minor: Read failed

error #002: ../../../src/H5Fsuper.c in H5F__super_read(): line 420: file signature not found
    class: HDF5
    major: File accessibility
    minor: Not an HDF5 file

1. In order to correct my mistake, I checked the following two guidelines-
2. https://github.com/basilkhuder/Seurat-to-RNA-Velocity

3. https://smorabit.github.io/tutorials/8_velocyto/

4. After a bit of research, I have summerized the process in following way- 1) Generate loom file from velocyto- by using following code- velocyto run -b filtered_barcodes.tsv -o output_path -m repeat_msk_srt.gtf bam_file.bam annotation.gtf

2) Get embeddings from Seurat (which I have already performed)

3) Merge loom files (needs to be performed separately for each sample used to create seuratobject) and Meta-data together in scvelo-

Then run RNA velocity.

Please let me know whether this procedure is correct. Thank you in advance.