when i use fq2bam_meth in docker, the log file shows Did not pass one of the seeding asserts on device: returning from GPU kernels early. how to solve it?
Hi wangyanni541,
Thanks for asking. So this is not a failing. It just tells you that the GPU cannot handle this data point so we move it to CPU to handle it. But the result stays valid.
Did the process finish properly? And do you get a valid bam output?
1 Like
yes, it is not a failing. I can get a bam file from the command, But this process is very memory and time consuming. And I discovered a pattern. Generally, single cell methylation data will encounter this problem. It seems that the reason is that too many chains are generated. My data was downloaded from this link （GSE215353 GEO Accession viewer.
Hi Wangyanni541,
Thank you so much for the information here. I think you are right, we haven’t optimized fq2bam_meth on single cell methylation data specifically. We will put this on our todo list and hopefully you will see the speedup sometime in the future.
I think it may be related to the number of kmers and random primers. When I merged multiple single-cell fastq files together, shuffled them, and then processed them with fastp, it seemed to pass the internal review of fq2bam_meth.
This tool is really excellent, it helped a lot. I hope you can update it as soon as possible. I look forward to it very much.