There are too few interaction signals to manually adjust them in Juicebox

[2benaszq]

Hello,

First of all, thank you very much for developing such an excellent phasing and scaffolding tool. However, I have encountered some issues while using it.

I first assembled hap1.fa and hap2.fa using hifiasm, each approximately 1 GB in size. Then, I merged them into a single file named output.fa. Subsequently, I used the following commands:

cphasing pipeline -f output.fa -pcd porec.fq.gz -t 100 -n 18:2 -hcr
ln -sf cphasing_output/porec.pairs.gz ./
ln -sf cphasing_output/4.scaffolding/groups.agp ./
cphasing pairs2mnd porec.pairs.gz -o porec.mnd.txt
cphasing utils agp2assembly groups.agp > groups.assembly
docker run -i --rm -w ${PWD} -u $(id -u):$(id -g) -v /calculate:/calculate -v /data:/data hic:v2.1 /software/3d-dna-201008/visualize/run-assembly-visualizer.sh groups.assembly porec.mnd.txt

I then loaded the resulting groups.assembly and groups.hic files into Juicebox for manual adjustment, but I found that the interaction signals were too sparse to make adjustments effectively.
Here is a screenshot of the Juicebox view:

Additionally, the scaffolding plot generated directly by cphasing (without manual adjustment) looks as follows:

The results seem excellent, and I only need to adjust a few contigs to achieve a nearly perfect phased genome. However, due to the low signal visibility in Juicebox, I am unable to complete this task.

As mentioned earlier, the merged genome is 2 GB in size, and the porec data is only about 25 GB. I wonder if the low data volume in phasing mode is causing the weak signal display in Juicebox, or if there is any way to enhance the signal visibility in Juicebox to help me adjust misassembled contigs. Alternatively, what amount of porec data would be required for this genome to display strong interaction signals in Juicebox

Thanks!

[wangyibin]

Sorry for my late reply,
Firstly, thank you for your feedback on this issue.

The heatmap plotting from cphasing shows that 25 Gb of pore-c is enough to adjust the genome in Juicebox.

Your genome’s high homozygosity may result in most of the mapping quality of pore-c fragments being smaller than 1.
By default, cphasing pairs2mnd removes the interaction with a quality <1; you can set -q 0 to load all the interactions of pore-c to the Juicebox.

cphasing pairs2mnd porec.pairs.gz -o porec.mnd.txt  -q 0

Best regards,
Yibin

[2benaszq]

Dear Dr. Yibin,

Thank you for your response.
With your help, I was indeed able to achieve the results I wanted.

However, I have one more question that I didn’t mention last time.
I noticed that CPhasing seems to call the program Partig, which appears to have a limitation on the contig length, requiring contig lengths not to exceed 2**27 bp, approximately 134 Mb. My genome, however, has a contig as long as 136 Mb. To address this, I manually split it into 100 Mb and 36 Mb segments, completed the scaffolding, and then merged them back together. While this approach works, it feels a bit inconvenient.

Is there any way to avoid this issue, or will this minor limitation be fixed in future updates?

Best regards,
Iron Man

[wangyibin]

Thank you for your suggestion.

We will fix this limitation in the future or add a function to split long contigs.

Best regards,
Yibin.

[2benaszq]

Dear Dr. Yibin,
Thank you for your reply, I look forward to the next update of CPhasing with great anticipation!

Best regards,
Iron Man

[2benaszq]

Dear Dr. Yibin：
I have recently been trying version 0.2.5.r291 of CPhasing. The work I am doing involves running CPhasing with Hi-C data and using a haploid mode for mounting because my genome only has one set of pseudo-haplotypes. After using Juicebox for the adjustment, I obtained the groups.review.assembly file. I wanted to use the adjusted results to generate an interaction heatmap again, so I ran the following commands:

cphasing utils assembly2agp groups.review.assembly -n 18
cphasing agp2fasta groups.review.agp genome.contigs.fasta > groups.review.chr.fasta
cphasing-rs pairs-break -o corrected.pairs.gz pass.pairs.gz groups.review.corrected.agp
cphasing pairs2cool corrected.pairs.gz cphasing_output/genome.contigs.contigsizes pass.q1.10k.cool -q 1 -bs 10k
cphasing plot -a groups.review.agp -m pass.q1.10k.cool -o groups.q1.500k.wg.png -bs 500k -oc

The resulting groups.q1.500k.wg.png is as shown below:

However, the interaction signals in Juicebox are very good, as shown here:

I am not sure where the issue lies—whether I made a mistake in my operations—but when I used the phasing mode previously, the generated plots didn’t seem to have any issues. I look forward to your reply.

Best regards,
Iron Man

[2benaszq]

Dear Dr. Yibin： I have recently been trying version 0.2.5.r291 of CPhasing. The work I am doing involves running CPhasing with Hi-C data and using a haploid mode for mounting because my genome only has one set of pseudo-haplotypes. After using Juicebox for the adjustment, I obtained the groups.review.assembly file. I wanted to use the adjusted results to generate an interaction heatmap again, so I ran the following commands:

cphasing utils assembly2agp groups.review.assembly -n 18 cphasing agp2fasta groups.review.agp genome.contigs.fasta > groups.review.chr.fasta cphasing-rs pairs-break -o corrected.pairs.gz pass.pairs.gz groups.review.corrected.agp cphasing pairs2cool corrected.pairs.gz cphasing_output/genome.contigs.contigsizes pass.q1.10k.cool -q 1 -bs 10k cphasing plot -a groups.review.agp -m pass.q1.10k.cool -o groups.q1.500k.wg.png -bs 500k -oc

The resulting groups.q1.500k.wg.png is as shown below:

However, the interaction signals in Juicebox are very good, as shown here:

I am not sure where the issue lies—whether I made a mistake in my operations—but when I used the phasing mode previously, the generated plots didn’t seem to have any issues. I look forward to your reply.

Best regards, Iron Man

Sorry, I misspoke earlier. It's still the porec data, but I'm using the --mode haploid. The command is as follows:

cphasing pipeline -f ../../00.data/genome.contigs.fasta -pcd pass.fq.gz -t 60 --mode haploid

[wangyibin]

I am very sorry for replying so late.

The problem of losing contacts is a bug of the cphasing pairs2cool command; you can add the parameter of --low-memory to generate a new .cool file to skip it. And we will fix it in the next release.

One other note: The pass.q1.10k.cool is generated from the corrected.pairs.gz, when you execute the plot you should replace the groups.review.agp to the groups.review.corrected.agp.

Best wishes.

[2benaszq]

Dear Dr. Yibin：
Thank you for your reply. The cphasing pairs2cool --low-memory parameter is indeed useful—I was able to obtain my desired results using it. However, when I tried to plot the final interaction heatmap with groups.review.assembly, it seemed that I still had to use cphasing plot -a groups.review.agp to generate the correct heatmap.

Since most of my genomes were manually fragmented during scaffolding, I was confused for some time about whether to use the groups.review.agp file or the groups.review.corrected.agp file for plotting. Taking the above case as an example, when I used the groups.review.corrected.agp file, the resulting heatmap was as follows:

It is obvious that the chromosome sizes do not match those in the Juicebox screenshot above.

However, when I used the groups.review.agp file, the heatmap appeared normal, as shown below:

So, should groups.review.agp indeed be used, or does the workflow of cphasing not align with your expected output?

Best wishes.
Iron Man