search

bioinfo-ibms-pumc / SCSA

SCSA: cell type annotation for single-cell RNA-seq data
GNU General Public License v3.0
82 stars 15 forks source link

Possible to accept scran::findMarkers results? #5

Open Lei-Guo opened 4 years ago

Lei-Guo commented 4 years ago

Most of people in our lab use scran::findMarkers to identify cluster markers. Is it possible for SCSA to accept the results from scran or at least let me know how to format scran's results for SCSA?

bioinfo-ibms-pumc commented 4 years ago

Thanks for the suggestion. I will test and add the function to accept the results from scran in the next modification.

bioinfo-ibms-pumc commented 4 years ago

I just added the function for scran and uploaded a test file, please see ReadMe file for detail. Thanks for the suggestion again.

Lei-Guo commented 4 years ago

Awesome! I'm recommending this tool to our lab.

Lei-Guo commented 4 years ago

Hi,

Thank you for adding support to Scran. However, I don't think it works correctly. "markers <- findMarkers(sce, sce$cluster)" returns a data frame of marker genes for each cluster. For example, if there are 12 clusters, the variable "markers" in the above code would contain 12 data frames. Each of the data frames contains the statistics for the marker genes of the corresponding cluster.

I don't think "python SCSA.py -d whole.db -s scran -i scran_pbmc_3k.csv -k All -g Human -p 0.05 -f 1.5 -b" returns correct results. If you look at the results, all clusters were annotated except cluster 1. Interestingly, "scran_pbmc_3k.csv" is the marker gene file only for cluster 1. I guess the program annotated the clusters based on logFCs, which is not right.

This is just my opinion. Please let me know what you think. Thanks. Regards

https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon Virus-free. www.avast.com https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

On Tue, Jun 23, 2020 at 12:40 AM bioinfo-ibms-pumc notifications@github.com wrote:

I just added the function for scran and uploaded a test file, please see ReadMe file for detail. Thanks for the suggestion again.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bioinfo-ibms-pumc/SCSA/issues/5#issuecomment-647903149, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACSH7P5RYYTFA2TOFBO27X3RYAW5DANCNFSM4OB4EEMQ .

-- Lei Guo

bioinfo-ibms-pumc commented 4 years ago

Dr Guo, Sorry for delay. Yes, absolutely this is my fault for the unfamiliar with Scran. Since Scran takes a whole pairwise comparisons for a cluster, not like other tools such as seurat which gives a comparison simply group other clusters together, it's hard to find a real LFC for annotation for the algorithm of SCSA. To solve the problem, I just use the pval.type="all" parameter to make sure a gene to be DE between each cluster and every other cluster. Then I calculate the mean LFC of all comparisons as the LFC of this cluster to do the further analysis. I had added additional R cmds in the ReadMe to do the task.

You may try to test the new sample "scran_pbmc_3k.csv" I just uploaded. And the predicted results is just fine. Please let me know whether it could be run normally. Also, any suggestions is welcome. Thank you for the correction again :)

Yinghao