An improved de novo assembly and annotation of the tomato reference genome using single-molecule sequencing, Hi-C proximity ligation and optical maps
Prashant S. HosmaniiD, Mirella Flores-GonzaleziD, Henri van de GeestiD, Florian MaumusiD, Linda V. BakkeriD, Elio G. W. M. SchijleniD, Jan C. van HaarstiD, Jan H. G. Cordewener, Gabino F. Sanchez‐PereziD, Sander PetersiD, Zhangjun FeiiD, James J. GiovannoniiD, Lukas A. MuelleriD, Surya SahaiD
Abstract
The original Heinz 1706 reference genome was produced by a large team of scientists from across the globe from a variety of input sources that included 454 sequences in addition to full-length BACs, BAC and fosmid ends sequenced with Sanger technology. We present here the latest tomato reference genome (SL4.0) assembled de novo from PacBio long reads and scaffolded using Hi-C contact maps. The assembly was validated using Bionano optical maps and 10X linked-read sequences. This assembly is highly contiguous with fewer gaps compared to previous genome builds and almost all scaffolds have been anchored and oriented to the 12 tomato chromosomes. We have found more repeats compared to the previous versions and one of the largest repeat classes identified are the LTR retrotransposons. We also describe updates to the reference genome and annotation since the last publication. The corresponding ITAG4.0 annotation has 4,794 novel genes along with 29,281 genes preserved from ITAG2.4. Most of the updated genes have extensions in the 5 and 3 UTRs resulting in doubling of annotated UTRs per gene. The genome and annotation can be accessed using SGN through BLAST database, Pathway database (SolCyc), Apollo, JBrowse genome browser and FTP available at https://solgenomics.net.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 289
- Organism
- Assay
- Research Focus
- Approach
In-depth summary
Claude Opus 5.5 · xhigh effort · 17 Oct 2026 · from the full text of version 1 (14 Sept 2019) · CC BY-NC-ND 4.0
A PacBio and Hi-C tomato genome rebuild shrinks gaps from 81.71Mb to 44Kb; the matching ITAG4.0 annotation adds 4,794 genes and updates most carried-over models.
Earlier tomato reference builds, assembled from 454 reads plus Sanger-sequenced BACs and fosmid ends, still had 81.71Mb of gaps and 3136 unplaced contigs in SL3.0, limiting genetic mapping, repeat analysis and annotation. SL4.0 is assembled de novo from long reads and scaffolded with Hi-C instead of BAC, FISH and genetic-map evidence, and comes with a re-annotation (ITAG4.0) that preserves legacy ITAG gene nomenclature.
Key results
ShownSL4.0 is a de novo PacBio assembly scaffolded with Hi-C into 12 chromosome-scale super-scaffolds; it is far more contiguous and has much less gap sequence than SL3.0 while remaining collinear with it.
Canu assembly of 80x PacBio reads: 504 contigs, 785 Mbp, contig N50 5.5 Mbp; 152 unplaced contigs form a 9.6 Mbp chromosome 00. Gaps given as 44Kb vs 81.71Mb in SL3.0. k-mer analysis shows one homozygous peak and 99.24% completeness; more than 36 Mbp of sequence missing from earlier builds is added, mostly repeats. · De novo assembly SL4.0; Pacbio data generation and de novo assembly; Introduction; Comparison of de novo assembled SL4.0 with…
ShownThe ITAG4.0 annotation has 34,075 protein-coding genes, carries over most ITAG2.4 models under their existing identifiers, adds 4,794 new genes, and agrees better with expression and homology evidence than ITAG2.4.
29,281 genes carried over from ITAG2.4, of which 21,962 were updated, mostly by 5' and 3' UTR extensions that double UTRs per gene. Genes with supporting evidence: 83% vs 76% in ITAG2.4; more than 80% of ITAG4.0 genes have annotation edit distance (AED) below 0.5, similar to Arabidopsis Araport11. AHRD assigns functions to 29,532 genes. · Updating legacy gene models and identifying…; Annotation Edit Distance as a metric of quality…; Migrating legacy annotation from ITAG2.4
ShownLong-read assembly recovers more repetitive DNA: REPET classifies a larger share of SL4.0 as repeats than of earlier builds, with LTR retrotransposons among the largest classes, more Gypsy elements and fewer unclassified repeats.
71.77% of SL4.0 is repetitive vs 60.94% in SL2.5 and 61.55% in SL3.0; 10% more Gypsy transposons than SL2.50; repeat regions up approximately 8% over SL3.0. The added sequence is largely repeat-rich, which the authors attribute to the limits of Illumina and 454 assembly. · Repeat Identification and Classification; Comparison of de novo assembled SL4.0 with…
ShownBefore the long-read build, the legacy assembly was upgraded to SL3.0 by integrating full-length BACs, correcting 454 homopolymer errors with Illumina reads, and fixing scaffold errors found with Bionano optical maps.
1,069 BACs removed 11,699,806 bases (11Mb) of contig gaps, with per-chromosome reductions of 3.17% to 49.07%; 16,723 homopolymer indels corrected. Optical map assembly: 674 cmaps, 762 Mb, N50 1.48 Mb, with 629 cmaps (93.3%) aligned. Two inversions on chromosome 12 were corrected and 19 gaps resized. SL3.0 is at NCBI (AEKE00000000.3). · Improvements in the legacy genome assembly; Bionano Optical Maps; BAC-based gap filling; Homopolymer correction
ShownNew genes are enriched for stress and defence functions, and the release adds lncRNA and gene-family resources, including an orthology analysis across 11 plant species.
Terms such as defense response (GO:0006952) and response to fungus (GO:0009620) are over-represented among novel genes (Fisher's exact test; values not in this text). FEELnc predicts 5,874 lncRNA genes with 6,694 isoforms. 348,442 of 409,806 genes (85%) fall in 7,445 orthogroups, yet tomato alone is given 15,933 families; these counts are hard to reconcile as printed. · Novel protein coding genes are involved in the…; Long non-coding RNA identification; Gene family analysis
Methods, data and reproducibility
PacBio reads (80x) were assembled with Canu, polished with Arrow and Pilon, and scaffolded with Hi-C (HiRise) into chromosomes, then checked against Bionano optical maps, 10X linked reads, k-mer spectra and BUSCO. Genes were re-annotated with MAKER from migrated ITAG2.4 models, Iso-Seq and public RNA-seq evidence plus Augustus and SNAP predictions.
- PacBio RS II and Sequel long reads of the reference cultivar Heinz 1706 · tomato (Solanum lycopersicum) · n = 80X coverage
- Illumina short reads used for Pilon polishing of SL4.0; Illumina reads provided by Syngenta were also used to correct 454 homopolymer errors in SL3.0 · tomato (Solanum lycopersicum) · n = 100X coverage for the Pilon set
- Hi-C library from young seedlings, generated by Dovetail Genomics · tomato (Solanum lycopersicum)
- Bionano Irys optical maps (Nt.BspQI labelling) and 10X linked reads used for validation · tomato (Solanum lycopersicum) · n = 674 cmaps
- SL3.0 legacy assembly, improved in this work with integrated full-length BACs · tomato (Solanum lycopersicum) · n = 1,069 BACs integrated · AEKE00000000.3
- PacBio Iso-Seq of breaker and mature green fruit · tomato (Solanum lycopersicum)
- Public wild-type RNA-seq from NCBI SRA (root, leaf, flower, callus, fruit stages) plus RenSeq NB-LRR and UTR-enriched libraries · tomato (Solanum lycopersicum) · n = more than 12 billion reads
- Phytozome proteomes of Oryza sativa, Arabidopsis thaliana, Vitis vinifera, Ipomoea nil, two Petunia, two Capsicum, Solanum tuberosum and S. pennellii, plus tomato · 11 plant species · n = 409,806 genes
- Methods & tools
- Canu v1.5, Arrow and Pilon (polishing), PBjelly and HiRise (scaffolding), Bionano IrysView 2.3, BUSCO, REPET, RepeatModeler, RepeatMasker, MAKER 3.1, Augustus (BRAKER1), SNAP, HISAT2, StringTie, Mikado, GMAP, AHRD, InterProScan, FEELnc, OrthoFinder, KinFin, CAFE
- Data & accessions
- https://solgenomics.net/
- Reagents & environment
- Versions are given for Canu v1.5 (with its full parameter list), IrysView 2.3, MAKER 3.1, GMAP 2017-06-20 and SRA tools 2.8.2-1, plus Pilon parameters; no containers or requirement files.
- Assay conditions & compute
- Only Canu memory and thread settings (e.g. minMemory=16G, batThreads=16) and Pilon with 32 threads are given; no hardware description or run times.
- Reproducibility gaps
- Genome and annotation are served through SGN (BLAST, SolCyc, Apollo, JBrowse, FTP), but the FTP link is missing from this extraction, no NCBI accession for SL4.0 or ITAG4.0 is printed, and analysis scripts are not mentioned.
Limitations (6)
- The BUSCO marker set was derived from the legacy ITAG annotation, so the authors note these builds naturally score well on it. · stated by the authors
- No gold-standard gene set exists for comparing annotations, and the authors still regard manual curation as the gold standard for gene models. · stated by the authors
- Stress-term enrichment among new genes may reflect the abundance of RNA-seq from fungus-treated plants and of NB-LRR-enriched RenSeq data in the evidence. · stated by the authors
- 152 contigs (9.6 Mbp) could not be placed on chromosomes and are concatenated into chromosome 00. · stated by the authors
- SL4.0 is described as validated with Bionano maps and 10X linked reads, but quantitative optical-map results are reported only for SL3.0, and BUSCO and mapping comparisons are in supplementary tables not in this text. · noted in this summary
- AED scores agreement with the same RNA-seq and protein evidence used to refine ITAG4.0, so better AED partly reflects the added evidence rather than independent validation. · noted in this summary
Who should read:Read this if you use the tomato reference genome or Solyc gene models and need to know what changed in SL4.0 and ITAG4.0, or want a worked example of upgrading a legacy plant reference with long reads and Hi-C.
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
- A high-continuity and annotated tomato reference genome
Xiao Su, Baoan Wang, … Tao Lin·BMC GenomicsPeer-reviewed·Dec 15, 2021
- Near-complete genome assembly of tomato (Solanum lycopersicum) cultivar Micro-Tom
Kenta Shirasawa, Tohru Ariizumi·Plant BiotechnologyPeer-reviewed·Jul 18, 2024
- A gap-free tomato genome built from complementary PacBio and Nanopore long DNA sequences reveals extensive linkage drag during breeding
W. M. J. van Rengs, M. H.- W. Schmidt, … C. J. Underwood·bioRxiv·Aug 30, 2021
- The de novo reference genome and transcriptome assemblies of the wild tomato species Solanum chilense
R. Stam, T. Nosenko, … A. Tellier·bioRxiv·Apr 26, 2019
- De novo genome assembly and transcriptome analysis for the drought and salt resistant Solanum sitiens
Corentin Molitor, Tomasz J. Kurowski, … Fady R. Mohareb·BioinformaticsPeer-reviewed·Jan 23, 2021
- Complete telomere-to-telomere assemblies of two sorghum genomes to guide biological discovery
Chuanzheng Wei, Lei Gao, … Yongfu Tao·iMetaPeer-reviewed·Apr 1, 2024