problem with running Kallisto on single cell data

Question

I am totally new to bioinformatics, therefore apologies if the terminology is not correct or the question is dumb :) I have been trying the following:

1. installed miniconda
2. create an environment called "rnaseq" where I installed kallisto conda install -c bioconda kallisto (0.50.1)
3. created another environment called "kb", where I installed Kallisto-bustools

conda create -y --name kb python=3.8
conda activate kb
pip install kb-python

4. run the following line of code from terminal:

kb count \
pbmc_1k_v3_S1_mergedLanes_R1.fastq.gz pbmc_1k_v3_S1_mergedLanes_R2.fastq.gz \
-i Homo_sapiens.GRCh38.cdna.all.index \
-x 10XV3 \
-g t2g.txt \
-t 8 \
--cellranger

I get this error message: ````kb_python.config.UnsupportedOSError: Failed to find compatible kallisto binary. Provide a compatible binary with the --kallisto option or run kb compile.

I have a MacBook Pro 2020, Intel Core i5, OS Sonoma 14.4. I used the same ".index" file and ran Kallisto on bulk RNA-seq data and it worked. I tried now to redownload the fastq files (they belong to one of the test datasets from 10x), regenerate the index file, remove all the conda envs and reinstall everything but nothing worked. Thanks for any help you could provide!

     

Answer 1

created another environment called "kb", where I installed Kallisto-bustools

This is known to work well. What about installing bustools in the same env as Kallisto?

Alternatively you can create a new conda environment

conda create --name kba
conda activate kba
conda install -c bioconda kallisto
conda install conda-build
conda skeleton pypi kb_python
conda build kb_python
conda install --use-local kb_python
kb count \
pbmc_1k_v3_S1_mergedLanes_R1.fastq.gz pbmc_1k_v3_S1_mergedLanes_R2.fastq.gz \
-i Homo_sapiens.GRCh38.cdna.all.index \
-x 10XV3 \
-g t2g.txt \
-t 8 \
--cellranger

If the above doesn't work you simply delete the env

conda deactivate
conda env remove --name kba

---

Okay skeleton didn't work. The last option is to use Docker. There is a Docker file of kallisto https://pachterlab.github.io/kallisto/download It's a bit weird it's not in the Docker repo. The Docker builder file is https://github.com/cyverse/docker-builds/blob/master/kallisto/Dockerfile

I assume most people know how to use a Docker file. If not, i.e. you want a tutorial, then please post as a separate question (quite alot of text to explain).

Answer 2

Well, it's embarassing to say but...I didn't have Autotools, CMake, and Make installed! Once I installed them, I could run the command. However, I am running now in another problem: even though the process starts, it gets stuck very quickly at the step "generating count matrix" (over night). But bustools is running, using 99% CPU but only 1 thread (even if I said "-t 8". My "counts_unfiltered" folder is of course empty. I never get an error message but I don't think that even on my modest laptop a dataset of 1K cells should take so long, or am I wrong?