Fig 1.
The Puzzle Hi-C pipeline contains three steps: mapping, scaffolding, and building. Ordering and orientation adopt an iterative method to obtain accurate assemblies via multiple iterations. Puzzle Hi-C introduces a dynamic, triangle window strategy during assembling. The triangle window is initially small but expands with interactions to produce more effective clustering. Finally, the genome is assembled according to the scaffolding results and output in final fasta and apg format files.
Fig 2.
Contact probability and strategies used to evaluate distance between scaffolds adopted by different software.
a, Ideally, the distribution of Hi-C contact is 1/x. Heat map shows the diagonal position of interaction density is very high, the farther away from the diagonal, the lower the interaction density becomes. b, Heat map shows the chr2 [0–35MB] assembled by LACHESIS, where c1, c2, c3 and c4 represent scaffolds and the rectangle represents the number of Hi-C reads links with two scaffolds. Compartment and TADs produce many long-range interactions with densities higher than adjacent interaction densities. c, Strategy to evaluate distance as adopted by LACHESIS and ALLHiC. d, Strategy to evaluate distance as adopted by 3D DNA. e, Strategy to evaluate distance as adopted by SALSA2. f, Strategy to evaluate distance as adopted by Puzzle Hi-C. g, CV of interaction density with the triangle region and square region. h, Errors of the distance between two scaffolds with different gap sizes in Puzzle Hi-C. i, Errors of different strategies to evaluate the distance between two scaffolds with 1000 samplings.
Fig 3.
Number of different errors generated by different software with different scaffold size.
a-c, inversions, relocations and translocations generated by LACHESIS, SALSA2, 3D DNA, ALLHiC, YaHS, and Puzzle Hi-C under different length of Scaffolds. d-f, Inversions, relocations and translocations generated by LACHESIS, SALSA2, 3D DNA, ALLHiC, YaHS, and Puzzle Hi-C with 25 sampling data under different length of scaffolds.
Table 1.
Number of errors generated by different software from human genome hg38 data split into different contig lengths.
Fig 4.
The synteny of chromosomes assembled by Puzzle Hi-C and LACHESIS compared with GRCh38.
Synteny between the assembly of chromosomes from scaffolders and GRCh38 chromosomes.
Table 2.
Number of errors generated by different software with human genome GCA_001013985.1.
Fig 5.
The synteny of chromosomes assembled by Puzzle Hi-C and other software compared with gayal chromosome 2.
Synteny between the assembly of chromosomes from scaffolders (LACHESIS, SALSA2, 3D DNA, ALLHiC, YaHS and Puzzle Hi-C) and manually curated chromosome 2 of gayal.
Fig 6.
The scaffolding results of LACHESIS and Puzzle Hi-C on T. bimaculatus.
a, Synteny between T. bimaculatus and T. rubripes; b, Synteny between Puzzle Hi-C-corrected T. bimaculatus and T. rubripes; c, i T. bimaculatus genome chr1 Hi-C heat map, the black box is the Hi-C heat map suggesting assembly error; ii T. bimaculatus genome chr1 Compartment, red is Compartment A and blue is Compartment B; iii The rearrangement of T. bimaculatus genome chr1 Compartment according to the corrected chr1; iv Puzzle Hi-C corrected chr1 Compartment after Puzzle Hi-C correction; v Puzzle Hi-C corrected chr1 Hi-C heat map.