Scaffolding orders and orients assembled contigs into chromosome-scale scaffolds using long-range linking data, inserting runs of N as gap spacers between joins. It adds no new sequence - so a scaffold N50 can be far larger than the underlying contig N50, and the two must be reported separately. Hi-C / Omni-C proximity ligation is the dominant modern modality (YaHS, SALSA2, 3D-DNA); genetic/linkage maps (ALLMAPS), Bionano optical maps, and reference homology (RagTag) are alternatives keyed to the data available. The single most important point: automated scaffolding produces a draft chromosome structure. It becomes trustworthy only after manual curation of the contact map (the VGP/Darwin Tree of Life standard), where misjoins, inversions, and false duplications are fixed by eye. Reference-guided scaffolding additionally imposes the reference's karyotype, erasing real fusions/translocations - so it must never be used on a genome whose chromosome structure is a biological question.
# Hi-C scaffolders
conda install -c bioconda yahs salsa2
# Hi-C read mapping + dedup
conda install -c bioconda bwa chromap samtools
# Reference-guided + linkage maps
conda install -c bioconda ragtag jcvi # jcvi provides ALLMAPS
# QC: telomeres, stats
conda install -c bioconda tidk seqkit assembly-stats
# 3D-DNA / Juicer (clone; juicer_tools is a jar)
git clone https://github.com/aidenlab/3d-dna.git
wget https://github.com/aidenlab/juicer/releases/download/v2.0/juicer_tools.2.0.jar
# Curation viewers (no conda; from GitHub releases)
# PretextMap / PretextView / PretextSnapshot (wtsi-hpag), Juicebox Assembly Tools (aidenlab)
Tell your AI agent what you want to do: - "Scaffold my draft contigs to chromosome level with Hi-C using YaHS" - "Map my Hi-C reads correctly before scaffolding (each end separately, dedup)" - "Prepare a .hic file and AGP so I can curate the scaffolds in Juicebox" - "Scaffold against a closely related reference, but only for gene coordinates" - "Order my scaffolds with a genetic linkage map using ALLMAPS" - "Check telomeres and report contig N50 vs scaffold N50 separately"
"I have HiFi contigs and an Arima Hi-C library. Map the Hi-C reads with each end aligned separately and PCR duplicates removed, run YaHS with the correct enzyme, and give me the AGP plus a .hic file for Juicebox curation." "My Hi-C is Omni-C (enzyme-free). Scaffold with SALSA2 using -e DNASE and -m yes so chimeric contigs get broken first, and tell me why -m yes matters."
"Render the contact map with PretextMap and walk me through what a misjoin, an inversion bowtie, and a leaked-haplotig stripe look like so I can decide what to break in PretextView." "The automated scaffolds_final looks chromosome-scale but a downstream synteny analysis disagrees - explain whether I am shipping an uncurated draft and what curation would change."
"Scaffold my non-model genome against a related reference with RagTag - but first tell me whether this will erase any real chromosome fusions, because I care about karyotype evolution." "I have two genetic maps. Integrate them with ALLMAPS to order and orient my scaffolds and use it to validate my Hi-C chromosome assignment."
"Report contig N50, scaffold N50, gap count, and total N bases separately, then run tidk to find which scaffolds have telomeres at both ends." "Gap-fill the scaffolds with my ONT reads, but only after curation, and flag any filled gap whose length is far from the estimate."
bwa mem -5SP or the per-end pipeline, or chromap), MAPQ-filter, and mark/remove duplicates.-m yes / YaHS contig error-correction on by default)._scaffolds_final is a first pass, not a finished genome - curate the contact map before trusting chromosome structure or claiming fusions/synteny.-e GATC for DpnII/Arima, omit -e (YaHS) or -e DNASE (SALSA2) for Omni-C/enzyme-free. Wrong enzyme modeling degrades joins.