Diploid Genome

[ziyi10]

Excellent software, I would like to know how to predict the whole-genome chromosomal structure of a diploid mouse using Hi-C data. How should I implement this? I ran the relevant features for three days without any output—could there be any issues?

pastis-poisson --counts ./test.npy --lengths 196 183 161 157 152 150 146 130 125 131 123 121 121 125 105 99 95 91 62 172 --ploidy 2 --outdir ./ --chromosomes chr1,chr2,chr3,chr4,chr5,chr6,chr7,chr8,chr9,chr10,chr11,chr12,chr13,chr14,chr15,chr16,chr17,chr18,chr19,chrX --seed 666

[gesinecauer]

Hey there! Thank you for your question.

Diploid inference can indeed be quite slow... especially when inferring a structure with 5,290 beads, since the size of the structure being inferred has a big impact on speed. If you are able to infer structures at a coarser resolution, that would definitely help.

Also, it looks like you aren't using any constraints, is that correct? I would definitely recommend using both constraints, with various combinations of penalties. Penalties can be specified with --bcc_lambda (for the 'bead chain connectivity' constraint) and --hsc_lambda (for the 'homolog-separating' constraint).

Unfortunately, there is currently no way to set the penalties automatically, so I recommend trying a variety of combinations and using the outputted structures to decide. If the 'homolog-separating' constraint penalty is not high enough, homologs will overlap or even be directly on top of one another in 3D space. If the 'bead chain connectivity' constraint penalty is not high enough, you'll notice large gaps between genomically neighboring beads (those that are adjacent on a molecule of chromatin).

[ziyi10]

Thank you for your reply, I ran the following results after adding --bcc_lambda and --hsc_lambda parameters

What I understand is that it has successfully run and generated three files:

Haploid seems to generate a pdb file. I wonder how to visualize the .coord file here.

What are the functions of the three name identical files generated in the folders "struct_draft_lowres" and "struct_draft_fullres"?

[gesinecauer]

Sure! We've provided two different examples of how one might easily visualize inferred structures with Matplotlib, which you can find here:

• Example of plotting multiple chromosomes together (with all chromosomes in the same plot)
• Example in which chromosomes are plotted separately (with each chromosome having its own plot)

If your visualization needs are more complicated than what can be achieved with Matplotlib, I suggest using Mayavi. Although it is fairly trivial to adapt our Matplotlib plotting example to use Mayavi instead of Matplotlib, Mayavi can be challenging to install. If you are interested in exploring Mayavi, I suggest you refer to their documentation and github for help with installation and usage.

You do not need to do anything with the files found in "struct_draft_lowres" and "struct_draft_fullres". These structures were generated to help configure the homolog-separating constraint - they are merely utilized to give a (very approximate) estimate for the distance between the centers of mass of homologous chromosomes.

[ziyi10]

output.log
Thank you for your answer! I found that when I set --bcc_lambda 1e8 --hsc_lambda 1e10, I could get the simulated .coords file as above, but there were indeed large gaps between genomically neighboring beads. Therefore, I adjusted the parameters to --bcc_lambda 1e10 --hsc_lambda 1e10. The output log is as above; it seems there were no errors, but no files were outputted. Why is this the case?