Link data not exhausted. fixable inversion and unlinked scaffolds

[talioto]

We have started to use Hi-C data in our projects and have been trying SALSA2 for several of them. For a particular genome (C-value 1.3Gb), it has 1% higher BUSCO than the 3D-DNA assembly. However, after looking at the contact maps in Juicebox we realized that some inversions could be fixed in the longest super-scaffold and others. Also, there remained many off-diagonal contacts that would indicate further joins to be made. So, it seems it did not converge to the most consistent or contiguous assembly.
There is still the option to increase iterations (has this been removed or not?) so we ran it with 10 instead of 3 iterations. However, it stopped at 3. Is there any way of increasing the number of iterations?

Is there any other parameter that could help to improve the assembly This is the commandline I´ve used: python /path/to/SALSA/run_pipeline.py -s 1340000000 -c 1000 -i 10 -a assembly.fa -l assembly.fa.fai -b EASI_03.1_2.hicup.bed -e GATC,GANTC -o out -m yes -p yes

We’ve run the scaffolding with SALSA three times, and although the three had the same parameters (and ended up performing 3 iterations) the N50s were different: 16Mb, 24Mb and 28Mb, respectively. Obviously, more contiguity is not always more correct, as it could be associated to more mis-joins, but is it normal to find such range of contiguity in the outputs? BTW, the N50 should end up being around 92Mb for this genome.
2020.10.07.22.36.45.SALSA2.HiCImage.pdf
2020.10.05.14.27.55.HiCImage.pdf
2020.10.05.14.37.58.HiCImage.fixed_in_JBAT.pdf

[ghuryejay]

SALSA shouldn't give different results with different input runs with the same parameters. Was it run to completion in all the runs?