henhaobio
Sign in

BioinformaticsBioinformaticsPeer-reviewed journal article

YaHS: yet another Hi-C scaffolding tool

Chenxi ZhouiD, Shane McCarthyiD, Richard Michael DurbiniD

PDFDOIPreprintCode
In-depth full-text digest available · Jump to digest ↓

Abstract

We present YaHS, a user-friendly command-line tool for construction of chromosome-scale scaffolds from Hi-C data. It can be run with a single-line command, requires minimal input from users (an assembly file and an alignment file) which is compatible with similar tools, and provides assembly results in multiple formats, thereby enabling rapid, robust and scalable construction of high-quality genome assemblies with high accuracy and contiguity. YaHS is implemented in C and licensed under the MIT License. The source code, documentation and tutorial are available at https://github.com/c-zhou/yahs.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (30 Jun 2022) · CC BY 4.0

YaHS, a C command-line Hi-C scaffolder, normalises inter-contig contacts by intra-contig medians, yielding more contiguous, accurate chromosome-scale scaffolds than SALSA2 and pin_hic.

Biodiversity genome projects such as the Darwin Tree of Life need chromosome-scale assemblies, but long reads alone rarely achieve this and existing Hi-C scaffolders (LACHESIS, HiRise, 3D-DNA, SALSA2, pin_hic) are sensitive to repeat content, library preparation and coverage. YaHS introduces a contact-matrix construction that normalises inter-contig signal by intra-contig expectations, aiming at more accurate joins and robustness to contig errors.

Key results

  1. ShownOn a simulated fragmented T2T-CHM13 human assembly, YaHS reconstructed far more contiguous scaffolds than SALSA2 and pin_hic, placing most sequence into chromosome-sized scaffolds.

    Input: 5,483 contigs, N50 715Kb. YaHS N50 132.6Mb (L50 = 9), N90 36.5Mb (L90 = 23), over 92% of sequence in 25 major scaffolds; SALSA2 N50 43.9Mb (L50 = 22), N90 3.3Mb (L90 = 103); pin_hic N50 51.5Mb (L50 = 19), N90 8.5Mb (L90 = 68). · 2.2 On simulated human genome assemblies; Figure legends

  2. SuggestedWhen aligned back to T2T-CHM13 with QUAST-LG, YaHS scaffolds from the error-free simulated assembly contained fewer misassemblies than those from either competing tool.

    Relocations/inversions: YaHS 40/21, SALSA2 262/55, pin_hic 171/115; no inter-chromosomal translocations reported for any of the three assemblies. · 2.2 On simulated human genome assemblies

  3. SuggestedAfter synthetic misjoins were added to the simulated input, YaHS found and broke most of them and kept its contiguity. SALSA2 fixed fewer errors, pin_hic became more fragmented, and QUAST-LG reported far more misassemblies for both than for YaHS.

    25 erroneous contigs with 30 errors (32.1Mb affected; 5,453 contigs, N50 718Kb). YaHS corrected 28 of 30 vs SALSA2 14 of 30; YaHS statistics unchanged, plus two relocations and two translocations. Relocations/inversions/translocations: SALSA2 278/18/63, pin_hic 188/22/123; pin_hic N50 fell to 37.5Mb (L50 = 28). · 2.2 On simulated human genome assemblies

  4. ShownFor the oak bush-cricket (Meconema thalassinum), a DToL genome estimated at over 9Gb, YaHS flagged hundreds of contig errors and produced gigabase-scale scaffolds. Its contiguity exceeded both SALSA2 and pin_hic.

    Input 2,093 contigs, 9,054Mb, N50 10.7Mb (L50 = 229); 268 errors detected. YaHS: three scaffolds longer than 1,349 Mb cover over 50%, 13 longer than 179.6 Mb cover 90%, largest 2,087Mb. SALSA2 N50 79.0Mb (L50 = 18), N90 3.5Mb (L90 = 311); pin_hic N50 202.4Mb (L50 = 12), N90 14.2Mb (L90 = 82). · 2.3 On Darwin Tree of Life assemblies

  5. ClaimedAcross many DToL species of varied taxonomy, genome size and starting quality, YaHS reportedly gave higher contiguity than SALSA2 and pin_hic, most clearly on L90.

    Stated for 15 DToL species; the per-species statistics are in Supplementary Table S1 and Figures S1-S15, which the provided text does not include. · 2.3 On Darwin Tree of Life assemblies; 3 Conclusion

Methods, data and reproducibility

From contigs plus a BED/BAM Hi-C alignment, YaHS can break contigs with low coverage. Over rounds of increasing chunk size, it scores contig pairs by inter-chunk contacts normalised to median intra-contig contacts at equal separation, then builds, prunes and traverses a scaffolding graph.

  • Simulated error-free assembly made by randomly cutting T2T-CHM13 into 100Kb to 1Mb contigs · human · n = 5,483 contigs (N50 715Kb)
  • Simulated assembly with 25 erroneous contigs carrying 30 introduced intra- and inter-chromosomal misjoins · human · n = 5,453 contigs (N50 718Kb)
  • T2T Arima Hi-C reads mapped to the simulated assemblies · human · SRX10230901-SRX10230903
  • Darwin Tree of Life genome assemblies with Hi-C data, spanning taxonomic groups, genome sizes and initial assembly quality · multiple species · n = 15 species
  • Oak bush-cricket initial contig assembly (ToLID iqMecThal1) · Meconema thalassinum (oak bush-cricket) · n = 2,093 contigs, 9,054Mb
Methods & tools
YaHS, SALSA2, pin_hic, Arima Genomics mapping pipeline, QUAST-LG, Juicebox, Arima Hi-C, T2T-CHM13 reference assembly
Software & environment
YaHS is implemented in C and released under the MIT License; it accepts BED or BAM alignments. No version numbers or containers are given for YaHS, SALSA2, pin_hic or QUAST-LG.
Reproducibility gaps
Algorithm details and per-species DToL statistics are in Supplementary Methods, Table S1 and Figures S1-S15, not the main text; the main text lists no accessions for the DToL assemblies or Hi-C data.
Limitations (3)
  • Errors are harder to detect when a short contig from one chromosome sits between nearby contigs of another. YaHS missed both misjoins in one such case (a 170Kb chr19 contig between chr10 contigs separated by a 1.2Mb gap), which left two extra relocations and two translocations. · stated by the authors
  • Only SALSA2 and pin_hic are benchmarked; LACHESIS, HiRise and 3D-DNA are cited but not compared. The claim that YaHS is fast has no runtime or memory figures in the main text. · noted in this summary
  • The simulated contigs are random 100Kb to 1Mb cuts with synthetic misjoins, which may not resemble repeat-driven breakpoints in real assemblies. DToL comparisons report contiguity only, with no reference-based accuracy check in the main text. · noted in this summary

Read this if you scaffold contig assemblies with Hi-C and want a single-command alternative to SALSA2 or pin_hic that includes contig error correction.

Citations
2,489
FWCI
143.88
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISGhurye '19Lin '24Obinu '23Zeng '24Ghurye '17Mergez '26
79% semantic matchBioinformatics
Jay Ghurye, Arang Rhie, Brian P. Walenz+5 more·PLoS Computational Biology·Aug 21, 2019· 723 citations