Modkit pileup - segmentation fault and performance issues in high-coverage regions

Start by reproducing the two modkit pileup failures with the supplied commands, including --modified-bases inosine, the high-coverage region, and the comma-separated --region value. Use the reported output to investigate the segmentation fault, low CPU utilization, and contig-missing error. Done means each case has a documented resolution or confirmed workaround, with performance behavior checked on the high-coverage region.

Dear ONT staff,

I am working with modkit to profile m6A and inosine in a dRNA-seq sample, and I am encountering a couple of issues.

First, when I try to use the --modified-bases parameter, I encounter a segmentation fault error, with no additional information regarding the cause of the issue.

This is the command I am using:

~/Software/dist_modkit_v0.6.1_481e3c9/modkit pileup
/path/to/aligned/bam
/path/to/output/bed.gz
--reference /path/to/fa.gz
--log /path/to/log
--threads 10
--modified-bases inosine
--region 1

This is the output:

parsing region 1
discarded 0 contigs with zero aligned reads
parsed 1 base modification(s). Base modifications other than 'A:17596' will be counted as 'N_other'.
adding single-base motif: 'A 0'
Segmentation fault (core dumped)

Furthermore, I have a couple of genomic regions covered by ~1M reads (due to high expression of specific transcripts), and this results in the code getting stuck, with only 2 out of 10 CPUs (according to top) being actively used.

This is the output:

parsing region 2
discarded 0 contigs with zero aligned reads
attempting to sample 10042 reads
Threshold of 0.69921875 for base A is low. Consider increasing the filter-percentile or specifying a higher threshold.
using general workers
93037146 B written to output: Test.bed.gz [4.78 MB/s]
[00:00:18] ###############------------------------- 88000000/242193529 genome positions 4,732,930.0021/s 33s
1222798 rows written
0 ~records errored

Could you recommend a workaround to avoid this issue and/or improve performance so that all available CPUs are effectively utilized? I also tried with the --high-depth --max-depth 100 options but it did not help.

Finally, I attempted to skip the two chromosomes with very high coverage. However, when I pass a comma-separated list of chromosome names to the --region option (as suggested in the documentation), I encounter a “contig missing” error.

Thank you very much for your support.

Best regards,

Mattia