The wrong depth when identifying the collapsed contigs

[fuxiangxin0703]

Hello, I am conducting the analysis of the plant autotetraploid genome using the C-Phasing software. When using C-Phasing for the identification of collapsed contigs, I used the following command:

cphasing-rs paf2depth hifi.align.paf.gz chromsizes.txt -w 5000 -s 1000 -o hifi.align.depth -t 100
cphasing collapse from-depth hifi.align.depth

I obtained the following identification results for collapsed contigs:

2026-06-11T11:13:04 [INFO] - Load /hifiasm_result_20251113/chromsizes.txt
2026-06-11T11:13:04 [INFO] - Load /hifiasm_result_20251113/hifi.align.paf.gz
2026-06-11T11:13:04 [INFO] - /hifiasm_result_20251113/hifi.align.paf.gz treated as gzip (flate2 MultiGzDecoder)
2026-06-11T11:28:05 [INFO] - Successful output coverage (window 5000bp, step 1000bp) to hifi.align.depth
[11:31:00] INFO Identified 7,554,469,736 bp contigs with collapse.py:228
read depth >= 141.36
INFO Identified 23,854,464 bp contigs with read collapse.py:236
depth < 8.08

And the result of plot.kde.plot.png shows that the average sequencing depth is 80× :

However, the actual size of this genome is 40Gb, and the HiFi data volume is approximately 800+Gb. Therefore, the actual depth should be 20×. I used the entire 40Gb of the genome to conduct the identification of collapsed contigs. Why did the c-phasing calculation come out to be 80×? Is this result available?
In addition, the collapsed contig identified by collapse is as large as 7.55Gb. This is much larger than what was observed in the HiC heatmap. Should I trust the results obtained through the collapse process or the results from the HiC heatmap instead?
Looking forward to your reply.

[wangyibin]

Hi,

Thanks for your detailed report.

In previous versions, C-Phasing conservatively selected the dominant peak from the depth distribution as the main peak. This works well for most assemblies but can fail when there is a large amount of collapsed sequence, as in your case. As a result, the estimated main peak may be much higher than the true sequencing depth.

We have addressed this in the latest version, v0.3.1.r311, which allows you to manually specify the main peak. Please try:

cphasing collapse from-depth --main-peak 20 hifi.align.depth

where 20 is the expected single-copy sequencing depth for your dataset.

After rerunning, please check whether the total size of the identified collapsed contigs is more consistent with your expectation and with what you observe in the Hi-C heatmap.

Please let us know the updated results. They will help us further improve the collapse detection module.

Best,
Yibin