Why is there a big difference between kallisto and cellranger count result？

[aizhimin]

Why is there a big difference between kallisto and cellranger count result？

kb count -i /mnt/data/ref/transcriptome.idx -g /mnt/data/ref/transcripts_to_genes.txt -t 8 -m 16G --mm --h5ad --filter bustools --cellranger --gene-names --verbose --overwrite -x 10XV2 -o output/SRR8315735 SRR8315735_1.fastq.gz SRR8315735_2.fastq.gz

kallisto：

cellranger：

[Yenaled]

Have you looked at the unfiltered results from kallisto?

[aizhimin]

Have you looked at the unfiltered results from kallisto?

@Yenaled The filtering criteria for both are the same.

[aizhimin]

@Yenaled this result is unfiltered 。

[aizhimin]

@Yenaled The results are come from this same sample: https://www.ncbi.nlm.nih.gov/sra/SRR8315735

[Yenaled]

Is that the result from the unfiltered_results folder? (Aka not the folder that is created by the --filter=bustools).

Can you provide more info? -- how the index was created (the command used and where you got the files). What's in the run_info.json file? What is the pseudoalignment rate? What is the path to the matrix/h5ad file you are loading in? What do the first few lines of the matrix files look like? What does the knee plot look like?

With 25 million reads, hard to imagine that the average cell only gets a count of 1000 even if only 2/3rds of those reads map unless some post-processing was done.

[aizhimin]

@Yenaled run_info.json : { "n_targets": 0, "n_bootstraps": 0, "n_processed": 25005657, "n_pseudoaligned": 1955219, "n_unique": 1095764, "p_pseudoaligned": 7.8, "p_unique": 4.4, "kallisto_version": "0.48.0", "index_version": 0, "start_time": "Fri Feb 17 09:45:50 2023", "call": "/root/miniconda3/envs/jupyter39/lib/python3.9/site-packages/kb_python/bins/compiled/kallisto/kallisto bus -i /mnt/data/ref/transcriptome.idx -o output/v2 -x 10X V2 -t 8 SRR8315735_S1_L001_R1_001.fastq.gz SRR8315735_S1_L001_R2_001.fastq.gz" }

Loading in the path: counts_ unfiltered.

The matrix file content:

You mean，can‘t add the --filter=bustools ?

[aizhimin]

@Yenaled Is it possible to lack the correct 5' barcode whitelist as the input for kallisto ？

[github-actions[bot]]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days