Dear all,
I am integrating multiple public scRNA-seq datasets and have reached the step where I need to distinguish aneuploid from diploid cells. I tested the copyKAT tool on the full dataset (~80,000 cells), and it took about 7 days to complete. I then followed the tutorial recommendation and ran it sample by sample on the same total number of cells, but it still required around 7 days.
Is there a way to run copyKAT more efficiently so that the analysis finishes within a few days rather than taking this long?
Alternatively, has anyone tried an approach where cells are first clustered, a representative subset of cells from each cluster is selected for copyKAT analysis, and then the resulting copy number labels are transferred from the annotated cells to the remaining cells?
•
written
by
mrashad
1 answer
Do you have access to a high-performance computing (HPC) system with multiple cores, hopefully at least 50? Essentially, I am recommending that you run your samples in parallel, perhaps using something like GNU parallel, Snakemake or NextFlow.
Also, each copyKAT process accepts a parameter known as "n.cores", which allows you to use multiple cores in order to calculate distances between cells. This should speed up computation of each sample to a great degree.
Do note that you have to multiply the number of samples you are running in parallel by the number of cores you are using for each sample, in order to get your overall core usage. So, imagine you are running 5 samples in parallel with 4 cores each, so you are in total using 5 samples x 4 cores = 20 cores overall. This is why I asked you whether you have a multi-core computer or HPC.
If you do not have access to a computational server with at least a decent degree of computing power, it is difficult to run tools like copyKAT in a reasonable amount of time.
Log in to answer this question.
-
Cells from normal sample were incorrectly classified as tumor cells using copykat
written by tujuchuanli
-
How do I infer the proportion of aneuploid cells using sequencing data?
written by Actinidia •
-
Experience with copykat
written by Nona •
-
How can one determine if a sample size is representative in scRNA-seq analysis?
written by fifty_fifty •
-
Differential expression design considerations
written by bioinf.questions •
-
The effect of heterozygosity on genome alignment
written by Iseez •
-
PCA value for integrated analysis in scRNA
written by ek699 •
-
How to randomly subset a specific number of sequences from a large fastq file?
written by j.thomas93 •
-
Seurat on both pooled and unpooled samples
written by Kristin Muench
-
When can I distinguish between an SNV and a sequencing error?
written by Gene_MMP8
