
==== Front
G3 (Bethesda)
Genetics
g3journal
G3: Genes | Genomes | Genetics
2160-1836
Oxford University Press US

38805698
10.1093/g3journal/jkae114
jkae114
Genome Report
AcademicSubjects/SCI01180
AcademicSubjects/SCI01140
First draft reference genome and annotation of the alternative oil species Physaria fendleri
https://orcid.org/0000-0002-0024-4940
Johnston Christopher R BioDiscovery Institute, Department of Biological Sciences, University of North Texas, Denton, TX 76201, USA

https://orcid.org/0000-0002-1057-4708
Horn Patrick J BioDiscovery Institute, Department of Biological Sciences, University of North Texas, Denton, TX 76201, USA

https://orcid.org/0000-0003-4696-1811
Alonso Ana Paula BioDiscovery Institute, Department of Biological Sciences, University of North Texas, Denton, TX 76201, USA

Ingvarsson P Editor
Corresponding author: Denton, TX 76201, USA. Email: crj013989@gmail.com
Conflicts of interest The author(s) declare no conflict of interest.

9 2024
28 5 2024
28 5 2024
14 9 jkae11411 4 2024
01 5 2024
08 8 2024
© The Author(s) 2024. Published by Oxford University Press on behalf of The Genetics Society of America.
2024
https://creativecommons.org/licenses/by/4.0/ This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

In the wake of increasing demand for renewable energy sources, plant-based sources including alternative oilseeds have come to the forefront of interest. Hydroxy fatty acids (HFAs), produced in a few oilseed species, are important chemical feed stocks for industrial applications. An integrated approach was taken to assemble the first draft genome of the alternative HFA producer Physaria fendleri (n = 6), an outcrossing species with high heterozygosity. Both de novo transcriptome assemblies and genome assemblies were produced with public and generated sequencing reads. Resulting intermediate assemblies were then scaffolded and patched with multiple data sources, followed by super-scaffolding onto a masked genome of Camelina laxa (n = 6). Despite a current lack of available resources for the physical mapping of genomic scaffolds of P. fendleri, topography of the genome with respect to repeat and gene content was preserved at the scaffold level and not significantly lost via super-scaffolding. Read representation, gene and genome completion statistics, and annotation results illustrated the creation of a functional draft genome and a tool for future research on alternative oil species.

Physaria fendleri
lesquerella
genome annotation
Illumina
hybrid assembly
Agricultural and Food Research Initiative (AFRI) 2021-67034-35137 2021-67013-3777 USDA National Institute of Food and Agriculture 10.13039/100005825 U.S. Department of Agriculture 10.13039/100000199 NSF 10.13039/100000001 1928147
==== Body
pmcIntroduction

Plant-based specialty oils and fatty acids (FAs) are increasingly used for industrial purposes to address the worldwide demands for bio-based products (Gunstone 2008; Lee et al. 2015). For example, hydroxy FAs (HFAs) are produced in the seeds of a few plant species, with widespread applications in industry such as anticorrosion coatings, nylon, plastics, paints, cosmetics, soaps, detergents, and fuel additives (Binder et al. 1962; Lee et al. 2015; Ohlrogge et al. 2018). The formation of HFA estolides via esterification with other FAs enhances their commercial value and utility with favorable cold flow properties at low temperatures (Cermak et al. 2006). Castor (Ricinus communis L.) seed oil contains 80–90% HFAs in the form of ricinoleic acid or C18:1-OH (denoted as number of carbons: number of double bonds) (Cocuron et al. 2014). Unfortunately, castor production in the United States is generally banned for commercial use due to the presence of the toxin ricin along with potent allergens capable of triggering asthma and other severe responses (Davison et al. 1983; Thorpe et al. 1988; Cocuron et al. 2014).

Physaria fendleri (2n = 12) is another HFA-producing plant species and a member of the Brassicaceae family native to the arid southwestern United States and northern Mexico. While its seeds accumulate the HFA lesquerolic acid (C20:1-OH), its genetic similarity to other common crops (e.g. canola and Camelina sativa) and model species (e.g. Arabidopsis thaliana) has led to increased research as a model for HFA production and potential domestic replacement for imported castor oil (Horn et al. 2016; Mandáková et al. 2017; Azeez et al. 2022; Cocuron and Alonso 2023). While castor can produce variable amounts of HFAs, it is generally universally reported to be higher than P. fendleri. Previous literature reports Chilean castor accessions produce 46–54% of seed weight as oil, and Indian accessions are reported to range from 38 to 55% seed oil, with a maximum of 90% of total seed FAs comprised C18:1-OH (Anjani 2012; Román-Figueroa et al. 2020). In contrast, P. fendleri generally only produces about 25% of its seed weight as oil, with a maximum of 60% FA comprising C20:1-OH (Barclay et al. 1962; Cocuron et al. 2014). Due to these limitations, metabolic engineering and breeding are necessary for the full introduction of HFAs derived from P. fendleri into industry (Salywon et al. 2005; Azeez et al. 2022). Such efforts must aim to improve total oil quantity and/or C20:1-OH content as suitable for the aforementioned industrial applications.

Although the genome of P. fendleri has been partially constructed to study interspecies variation, a genome of high enough contiguity and precision for structural and thus functional annotation is lacking (Kiefer et al. 2019). The high frequency of outcrossing of this species results in additional challenges due to great sequence variation among individuals, especially when using integrated evidence. With advances in the integration of various sequencing read types, reference-directed genome assembly methods, and strategies for managing intraspecies/intrasample sequence variation, a collection of diverse data types can be effectively used for constructing a high-confidence genome assembly based on established workflows (Roach et al. 2018; Sohn and Nam 2018). These workflows include scaffolding and gap-filling tools that can not only make use of paired or long DNA reads but also RNA-seq reads of varying lengths (Xue et al. 2013; Song et al. 2016). The closing and filling of gaps can also be done using preassembled contigs, which is helpful to obtain higher confidence by using multiple assembler algorithms (Kosugi et al. 2015; Xu et al. 2020). Newer validated programs are able to correct potential misassemblies and rescaffold based on whole-reference genomes, including those of closely related species; this can then be followed by placing of scaffolds to pseudochromosomes using synteny and homology, producing validated references without optical mapping or Hi-C reads (Tamazian et al. 2016; Alonge et al. 2022). Finally, in addition to analyzing contiguity and read representation, genomes (and associated transcriptomes) can be carefully evaluated from multiple perspectives including expected ortholog content, completeness of coding regions, and synteny comparisons with related species (Reese et al. 2014; Reuscher et al. 2018; Manni et al. 2021). These processed genome drafts, along with publicly available sequence data of all types, are then effective inputs for gene prediction using canonical approaches employed by the MAKER pipeline, namely Augustus and SNAP (Korf 2004; Stanke et al. 2006; Campbell et al. 2014b; Hoff and Stanke 2019).

The objective of this work was to assemble a structurally annotated draft reference for P. fendleri by combining short paired-end genomic and RNA-seq reads, long RNA-seq reads, and preassembled contigs, including experimental and publicly available data. The functional annotation of this reference data set should provide oilseed researchers with a new tool for analyzing HFA metabolism and identifying potential targets for metabolic engineering. Finally, the strategy used for the construction of this draft aims to address some of the challenges of working with outcrossing plant species, such as the erroneous creation of duplicated regions and the inability to resolve haplotypes (Dierig et al. 2012; Roach et al. 2018).

Materials and methods

Plant materials and nucleic acid extraction

P. fendleri embryos were cultured in vitro in accordance with Cocuron and Alonso (2023), with minor modifications, to reduce variance in environmentally induced gene expression and normalize RNA yield. Embryos used for culturing were collected from mother plants of the WCL-YS1 (USDA accession: PI 610492) cultivar at 21 days after pollination, with embryos being excised from the seed coat prior to use (Dierig et al. 2000). Mother plants were grown from seed in a growth chamber set to 22°C, 320 µE m−2 s−1, and 50% relative humidity with a 16-h photoperiod. Culture medium consisted of 40 mM glucose, 5 mM glutamine, 2.0 µl ml−1 of 1,000× Gamborg vitamin solution, 20 mM HEPES, 2.2 mg ml−1 of Murashige & Skoog basal salts, 0.02 mM of (+/−) abscisic acid, and 23% (w/v) polyethylene glycol. This media resulted in a production of FA quantity and composition similar to that of the in planta embryos, approximately 25% w/w of total oil of which ∼55% was lesquerolic acid, when cultured for 9 days (Supplementary Tables 1 and 2). Cultures were maintained in an incubator set to 12 µE m−2 s−1 at 22°C and 50% relative humidity.

RNA was extracted from cultured P. fendleri embryos using a 2.0% w/v hexadecyl(trimethyl)azanium bromide (CTAB) extraction buffer containing a 2 M concentration of sodium chloride, 100 mM of 2-amino-2-(hydroxymethyl)propane-1,3-diol dihydrochloride (Tris-HCl) at pH = 8, 25 mM of 2-[2-[bis(carboxymethyl)amino]ethyl-(carboxymethyl)amino]acetic acid (EDTA) at pH = 8, and 3.0% w/v 1-ethenylpyrrolidin-2-one (PVPP) according to Johnston et al. (2022). After adding 2-sulfanylethan-1-ol and N1-(3-aminopropyl)butane-1,4-diamine (spermidine) to the buffer in order to achieve a concentration of 3.0% v/v and 0.5 g L−1, respectively, the buffer was heated to 65°C prior to extraction. Two sequential extractions were performed with 600 and 450 µL per sample. Following each extraction, a 24:1 (v/v) trichloromethane:3-methylbutan-1-ol solution was added and aqueous layers were combined in tubes on ice following centrifugation. This aqueous phase was further subject to phase separation with 1 equal volume of the 24:1 (v/v) trichloromethane:3-methylbutan-1-ol solution, with aqueous phase being collected in tubes on ice. Ethanol precipitation was carried out on the final aqueous extract by adding 0.10 volumes of sodium acetate, 0.25 volumes of RNA-grade glycogen (0.5 mg mL−1), and 2 volumes of ethanol prior to incubation at −20°C for 12 h. Following precipitation, extracts were centrifuged for 1 h at 13,000 × g and 4°C and RNA pellets were rinsed 3 times with 1 ml of 70% (v/v) ethanol and dried under a sterile bench. Samples were resuspended in RNase-free water and 0.3 volumes of 8 M lithium chloride were added prior to incubation at −20°C for 12 h. An additional 3 rinses with 70% (v/v) ethanol were carried out after centrifuging for 1 h at 13,000 × g, following which RNA pellets were resuspended in RNase-free water.

Library preparation and evidence sources

An integrated methodology using both generated RNA-seq reads and available genomic and RNA-seq reads from public databases (described below) was undertaken to produce a stronger consensus assembly (Fig. 1). For generated RNA-seq data, extracted samples were analyzed using capillary electrophoresis to determine RNA integrity and detect any contamination (Supplementary Fig. 1). Six replicates were used with approximately 500 ng of RNA for each replicate. The total library of 2 × 75-bp reads was prepared using the Illumina Stranded Total RNA Prep workflow and the NextSeq MidOutput (Illumina, San Diego, CA, USA) kit. Paired-end reads were generated and subjected to FastQC for quality control prior to usage (Sequencing Read Archive [SRA]: SRP430430). Additional P. fendleri read data from the NCBI SRA (https://www.ncbi.nlm.nih.gov/sra) was used for both genomic and transcriptomic assemblies. For the transcriptome assembly, RNA-seq data from single-end Illumina reads (BioProject: PRJNA291630; SRA: SRP061893) and 454 long reads (BioProject: PRJNA260225; SRA: SRP046070) were also integrated into the assembly process as described below (Kim and Chen 2015; Horn et al. 2016). For the genomic assembly, paired-end Illumina reads (BioProject: PRJEB26555; SRA: ERP108550) and the associated fragmented genome assembly of P. fendleri (GenBank Acc.: GCA_900406525.1) were used as described below (Kiefer et al. 2019). These genomic reads were run in the MaSuRCA assembler (v4.1.0) with k-mer coverage ranging from k = 49 to k = 75 to produce super-reads for additional contig evidence (Zimin et al. 2013).

Fig. 1. Summary of genome and transcriptome assembly workflow. Specifics about each step can be found in Supplementary Figs. 2 and 3.

De novo transcriptome assembly

A de novo transcriptome assembly was prepared by utilizing multiple assembler programs, k-mer sizes, data sets, and filtering algorithms (Supplementary Fig. 2). The effect of k-mer size (Supplementary Fig. 3) was first observed using SOAPdenovo-Trans at k = 25, 27, 29, 31, 33, 35, 37, and 39 with the aforementioned paired-end library (Xie et al. 2014). To ensure the inclusion of assembled transcripts detectable with a certain k-mer size, all 8 k-mer sizes were merged using the “transabyss-merge” function in Trans-ABySS (v2.0.1) (Robertson et al. 2010). Additional assemblies were performed on paired-end read data using rnaSPADES (“--ss rf” for strand-specific reads), Trans-ABySS (default settings), and Trinity (“--SS_lib_type RF”) (Robertson et al 2010; Grabherr et al. 2011; Bushmanova et al. 2019). To improve assembly quality, the public RNA-seq data (the single-end Illumina reads and the 454 long reads) were also analyzed separately through each assembler and merged, tuning parameters as necessary to account for the different read types.

Output from all 4 assemblers and all 3 data sets was merged using transabyss-merge, followed by an initial removal of duplicates via the “tr2aacds” function from EvidentialGene (http://arthropods.eugenes.org/EvidentialGene/) and one pass through CD-HIT-EST at an identity cutoff of 0.98 (Li and Godzik 2006). The resulting draft assembly was then passed through a modified version of the Bellerophon pipeline, in which reads were mapped to the draft with HISAT2 and transcripts comprising a sum of under 1 transcript-per-million across all samples were removed, followed by another run of CD-HIT-EST with an identity cutoff of 0.95 (Huber et al. 2004; Kim et al. 2015). Finally, chimera removal was carried out using the method developed by Yang and Smith (2013). For transcriptome assembly evaluation, the “brassicales_odb10” database was used since it was considered most precise for protein-coding gene evaluation in a Brassicaceae (Manni et al. 2021). All associated ORFs were detected using TransDecoder (v5.5.0) (https://github.com/TransDecoder/TransDecoder), integrating homology detected via UniProt and Pfam matches in order to produce predicted amino acid sequences (Bateman et al. 2023; Mistry et al. 2021).

Organellular genome assembly

Organellular genomes were constructed from the aforementioned genomic reads (Supplementary Fig. 4). The chloroplast genome was assembled at k = 85 using GetOrganelle (Jin et al. 2020). The mitochondrial genome was assembled using SPADES with k = 55 after testing from k = 33 to k = 95 (Supplementary Table 3) (Bankevich et al. 2012). The initial k = 55 assembly was first scaffolded using the genomic super-reads from MaSuRCA in LRSCAF and then was scaffolded around exonic regions by using RNA-seq data, with L_RNA_scaffolder (Xue et al. 2013; Qin et al. 2019) and P_RNA_scaffolder for the public long-read RNA-seq data and generated paired-end Illumina RNA-seq data, respectively (Zhu et al. 2018). Further scaffolding was carried out within protein-coding regions via PEP_scaffolder, using the predicted amino acid sequences from the de novo transcriptome (Zhu et al. 2016). A second de novo assembly was performed using GetOrganelle with k = 85 (Xu et al. 2020). This second de novo mitochondrion assembly was merged with the scaffolded SPADES assembly via MAC.

Nuclear genome assembly

The existing fragmented assembly was then used as an initial foundation for nuclear genome construction (Supplementary Fig. 4). The associated genomic reads were initially used to create secondary assemblies using 3 different assemblers: ABySS at k = 66 and k = 72, SOAPdenovo (r242) at k = 63 and k = 71, and MaSuRCA with auto-k (converged on k = 67) at k = 49 and k = 75 (Supplementary Table 4) (Luo et al. 2015; Jackman et al. 2017). All secondary assemblies along with the fragmented genome assembly were merged with MAC and then subjected to 3 rounds of scaffolding and gap filling with both genomic and transcriptomic data, each round being separated by a run in Purge Haplotigs to deal with the high heterozygosity (Supplementary Table 1) (Roach et al. 2018).

Following scaffolding and gap filling, the resulting 24,235 contigs were placed into pseudochromosomes according to synteny with the 6 chromosomes (same n as P. fendleri) of the Camelina laxa assembly (GenBank Acc.: GCA_024034495.1) using the Chromosemble tool from Satsuma (Grabherr et al. 2010; more information in Supplementary Table 3). The C. laxa target assembly was first masked according to the following. Miniature inverted-repeat transposable elements (MITEs) were detected using MITE-Tracker and combined with long terminal repeats (LTRs) detected according to Campbell et al. 2014b). More evolutionarily recent LTR retrotransposons were detected at a similarity 99% and were combined with older LTRs detected with a reduced similarity of 85%. Insertions nested within LTR elements were identified and likely false-positive elements in internal LTR regions were masked. The resulting LTR and MITE libraries were combined with repeats identified by RepeatModeler, and subsequently run against a database of known A. thaliana proteins (GenBank Acc.: GCA_000001735.2) manually cleared of transposable elements using BLASTx, and these BLAST hits were excluded from the repeat library (Camacho et al. 2009; Flynn et al. 2020). The combined repeat library was then used for masking the target C. laxa assembly prior to contig placement. The resulting assembly was subjected to 1 last round of scaffolding and gap filling, primarily to join unplaced contigs, resulting in 90.4% of the final 272,649,109-bp assembly residing on the first 6 pseudochromosomes.

Annotation

A limited functional annotation was carried out on the de novo transcriptome assembly using the Trinotate pipeline, following which only final sequences were kept for later usage by the MAKER pipeline (Campbell et al. 2014a, 2014b; Bryant et al. 2017). Prior to annotation, the nuclear genome was first scanned for sequence repeats according to the same methodology described above for super-scaffolding. Following this, the genome assembly quality was also evaluated by the LTR assembly index as described by Mokhtar et al. (2023). Gene prediction and annotation of the nuclear genome were carried out via 3 rounds of the MAKER pipeline. For the first round, the de novo transcriptome assembly and associated protein predictions were used as mRNA and protein homology evidence, respectively, and the combined repeat library was used for masking (Supplementary Table 5). Specifically, homology was detected by using BLAST hits against the reference proteomes (via BLASTp) of A. thaliana (Refseq Acc.: GCF_000001735.4), C. sativa (RefSeq Acc.: GCF_000633955.1), Brassica rapa (RefSeq Acc.: GCF_000309985.2), Raphanus sativus (RefSeq Acc.: GCF_000801105.2), and Brassica napus (RefSeq Acc.: GCF_020379485.1). A reference-free run of rnaQUAST was used with GeneMarkS-T to estimate the number of genes giving rise to the de novo transcripts assembled (Tang et al. 2015; Bushmanova et al. 2016).

In the first round, ab initio predictions were directly inferred from the provided mRNA and protein sequence information, following which predictions were exported to SNAP and Augustus for training hidden Markov models (HMMs) (Korf 2004; Stanke et al. 2006; Hoff and Stanke 2019). The HMMs produced by SNAP and Augustus were fed back into a second round of MAKER for refining gene predictions, on which SNAP and Augustus were retrained for a second iteration, finally producing input predictions for the third round of MAKER. The chloroplast genome and mitochondrial genome were annotated using GeSeq and Mitofy, respectively (Alverson et al. 2010; Tillich et al. 2017).

Synteny analysis

Analysis of interspecies genomic synteny within the P. fendleri assembly was carried out on both A. thaliana and C. laxa (GenBank Acc.: GCA_024034495.1) genomes using the Satsuma2 package (https://github.com/bioinfologics/satsuma2) with a minimum block size of 10 (Berardini et al. 2015). MCScanX was also used to examine collinearity within the constructed P. fendleri genome based on amino acid sequence alignments with BLASTP at E-value threshold of 1 × 10−5 (Wang et al. 2012). For interspecies synteny, syntenic blocks contained a minimum identity of 85.0%, a minimum length of 2009 bp (the average transcript length from the MAKER output), and a maximum gap size of 1 Mb for joining blocks on the same strand. For collinearity within P. fendleri, syntenic blocks had a minimum length of 1 Mb, and blocks were joined on the same strand within pseudochromosomes using a maximum gap size of 1 Mb. Syntenic blocks were not considered if the query and target sequence length differed by more than 3-fold.

Results and discussion

Transcriptome assembly performance

A total of 386,008,498 paired-end reads were generated for the transcriptome assembly. As no contiguous P. fendleri assembly of reference quality was available, only de novo transcriptome assembly was performed. To converge on a consensus assembly, a factorial combination of 4 assemblers and 3 different types of reads were used and evaluated by read representation (defined as the overall alignment rate of all reads used against the final assembly), the total number of transcripts, and a rough estimate of the number of genes predicted to give rise to those transcripts. Upon evaluation of the impact of k-mer size, all combinations of RNA-seq reads tested clearly displayed a preference for k = 31. However, this was only evident when paired-reads were analyzed alone and when evaluated by the gross number of total sequences yielded (Supplementary Fig. 2). The addition of new read types buffered the effect of k-mer size; however, based on the results, it can be seen that overall, read type as well as the assembler used had a much large effect on the number of sequences assembled and the mean sequence length. Since a robust filtering procedure was used later on and the effect of k-mer size was modest, all tested k-mer sizes were merged via transabyss-merge to form the representative SOAPdenovo-Trans assembly.

Although all assemblers were able to produce assemblies with >88.0% read representation when using only paired-end reads, the magnitude of the effect of adding new read types differed across assemblers despite a consistent trend within each assembler (Fig. 2). Trinity handled sequential addition of other read types the most effectively, only losing 7.0% read representation when adding single-end Illumina reads, and recovered to a total loss of only 3.5% read representation when long RNA-seq reads were added to both. Trans-ABySS produced the assemblies with the lowest read representation regardless of read type. It was found that merging the assembly created with all 3 read types, from each assembler, produced a set of 199,563 transcripts (of which only 68.0% were over 500 bp), highlighting the necessity for further processing (Supplementary Table 6). While 86.6% of expected BUSCOs were found complete, 75.9% were duplicated, which is likely high for a diploid Brassicaceae considering the reference-quality A. thaliana TAIR10.1 assembly contains 43.1% complete BUSCOs as duplicates (Berardini et al. 2015). Regardless, a high read representation of 93.6% was obtained from the merged, unfiltered assembly.

Fig. 2. Organellular genome annotations of P. fendleri. a) Chloroplast genome (Genbank Acc.: BK063463). b) Mitochondrial genome (GenBank Acc.: OQ818196).

To remove bulk fragments/artifacts and duplicated transcripts, the prefilter chosen, tr2aacds followed by CD-HIT-EST, reduced the total transcript count to 70,840 and the number of duplicated complete BUSCOs to 58.1% (Supplementary Table 7). A total of 73.9% of the BLASTX hits of the prefiltered assembly to TAIR10.1 protein products were full length (>80.0% query cover). The 3-step final processing procedure was then performed for a more precise removal of duplicates, fabricated transcript isoforms, and chimeras that was not subject to overpowering by the noise of bulk fragments and artifacts removed by prefiltering. Following all 3 steps of final processing, a final de novo transcriptome was yielded containing 74,416 total transcripts, 91.4% read representation, with 77.9% of BLASTX hits considered complete (>80.0% query cover), and 87.8% of expected BUSCOs complete, with 62% duplicated. The mean sequence length of the final assembly was 1,661 bp, comparable to that of the transcriptome of A. thaliana (1,663 bp) (Berardini et al. 2015). Following annotation with the Trinotate pipeline, the assembly increased to 75,096 transcripts, assumed to be additional splice variants detected by the TransDecoder step and homology. The transcriptome assembly was considered an independent piece of evidence for arriving at the draft P. fendleri genome, having been performed de novo.

Chloroplast and mitochondrial genomes

The chloroplast genome (Genbank Acc.: BK063463) was assembled with 391× coverage, and produced a single closed contig without unplaced scaffolds (Fig. 3). The accuracy of the chloroplast assembly was validated by >99.7% identity with the chloroplast of Physaria pinetorum (GenBank Acc.: MK637778.1). The mitochondrial genome (GenBank Acc.: OQ818196) was assembled with 262× coverage; however, the SPADES assembly alone did not produce a useable contiguous sequence even with the optimal k-mer size and further scaffolding (35 contigs in best assembly). Merging the additional GetOrganelle assembly with the SPADES one using MAC, 8 contigs were achieved; one last run of LRSCAF, L_RNA_scaffolder, and P_RNA_scaffolder as previously described produced a final contig of 255,307 bp (Fig. 3) (Tang et al. 2020).

Fig. 3. De novo transcriptome assembly evaluation. a) Effect of assembler and read combinations on read representation, defined as the overall alignment rate of generated reads that mapped back to the produced assembly. Paired-end = SRA: SRP430430; single-end = SRA: SRP061893; long reads = SRA: SRP046070. b) Effect of assembler on total transcripts produced, and total corresponding genes giving rise to the transcripts as predicted by GeneMarkS-T.

Step-wise genome assembly performance

The SRA genomic reads were reassembled and merged using a similar strategy to that of the transcriptome assembly; multiple assemblers were used; however, for each, 2–3 k-mer sizes were used. While ABySS had the best performance in terms of N50, it still only had a maximum N50 of 2,783 at k = 66. The same assembly had the lowest contig count (n) of all, but it was still highly fragmented with 140,873 contigs. Due to this fragmented pattern of assembly output, an integrated approach was taken to deal with scaffolding and eventual gap filling. Merging all de novo assemblies into a consensus assembly via MAC reduced the total n to 103,241 and achieved an N50 of 6,982. A first round of scaffolding and gap filling yielded an assembly with N50 of 9,871 and n of 89,502; in comparison, the fragmented P. fendleri assembly had only ∼10.0% more contigs, indicating an issue with the de novo assemblies.

As P. fendleri is an obligate outcrosser, it was expected to have a significant amount of haplotigs, likely inflating the true size of the genome (Dierig et al. 2012; Olson et al. 2015). This was determined to likely be the case, as the estimated genome size using the k-mer distribution with ntcard (Supplementary Fig. 5) indicated a mean k-mer coverage of 18.8, heterozygosity of 2.4%, and an estimated total genome length of 324 Mb (k = 26) (Mohamadi et al. 2017). Similarly, the fragmented assembly available online was 331 Mb. As a result of suspected haplotig content, the assembly was subjected to a run of Purge Haplotigs at a low read depth cutoff of 2, mid-point of 29, and high cutoff of 190, with a default 70.0% cutoff for assigning a contig as a haplotig. Remarkably, the n of the assembly dropped to 34,718, with an N50 of 14,432, and the total size of the genome was reduced to 269 Mb (Fig. 4). The final 2 rounds of gap filling, split by another precautionary run of Purge Haplotigs (low cutoff of 2, mid-point of 25, and high cutoff of 172 in accordance with the read depth histogram), was carried out. Only modest improvements were observed in terms of total contig count, N50, and gaps with each successive step in the final 2 rounds; however, following completion, a scaffold-level assembly with n of 24,235, N50 of 33,600, and 267 Mb in size was produced. The N90 at this level of the assembly was 4,563 bp, while the average transcript length in A. thaliana is 1,663 bp; due to this contiguity in gene-coding regions (BUSCO “embryophyta_odb10”: 77.3% complete) and the low count of total bases in gaps throughout the assembly (1.7%), contigs were subsequently super-scaffolded (Berardini et al. 2015). Concerning read representation, 84.0% of the paired-end RNA-seq reads and 78.2% of the genomic reads from the NCBI SRA mapped back to the final assembly (Table 1). This slight drop in representation of genomic reads is likely a result of haplotig content.

Fig. 4. Quantitative evaluation of genome size at each processing step in assembly workflow. The left axis corresponds to the base pair in the 6 longest contigs, as well as the total base pair in the assembly including unplaced scaffolds. The right axis corresponds to the N50 at each step. See Supplementary Table 3 for details on each step.

Table 1. Final assembly results, including 6 pseudochromosomes, mitochondrial assembly, and chloroplast assembly.

Assembly statistics	N50	41,908,749	
Total bp	246,881,883	
Total contigs	8	
RNA-seq read rep.a	83.95%	
DNA read rep.b	78.19%	
Gaps	87248	
% N	3.92	
Annotation resultsc	Gene models	39,859	
Avg. gene length	2,009.39	
% genes AED < 0.5	0.873	
Complete BUSCO hits	77.10%	
a RNA-seq read representation, with reads mapped via bowtie2 “--local”. SRA: SRP430430.

b DNA read representation, with reads mapped via bowtie2 “--local”. SRA: ERP108550.

c Annotation results correspond only to 6 pseudochromosomes, chloroplast, and mitochondrial genomes.

The final assembly contained a BUSCO completeness of 77.1% when including the chloroplast and mitochondrial genomes, but removing unplaced scaffolds (Table 1); this being so close in comparison to the BUSCO completeness of the fragments prior to super-scaffolding (n = 24,235) confirms that placement into the masked C. laxa genome did not distort or lose significant protein-coding gene structure (Table 2). Of the genes annotated by MAKER, 87.3% had an Annotation Edit Distance of under 0.5, comparable to well-annotated de novo genomes including those of diploid plants (Eilbeck et al. 2009; Campbell et al. 2014a; Sork et al. 2016). Gene density was shown to be generally well correlated with RNA-seq read representation on a global scale (Fig. 5).

Fig. 5. Genomic topography of the 6 pseudochromosomes of P. fendleri. Plot created using Circos (Krzywinski et al. 2009). a) Read depth of paired-end RNA-seq reads (SRA: SRP430430) in 200,000-bp intervals, maximum of 500 bp shown. b) Coverage of genomic paired-end reads (SRA: ERP108550) in 200,000-bp intervals, maximum of 40× shown. c) Repeat density, represented as the percentage of bases in identified repeats for each 200,000-bp interval. d) Gene density, represented as the number of identified protein-coding genes per 200,000 bp. e) GC content for each 200,000-bp interval. The inside line next to f) represents 0%, and the maximum is 66%. f) Inter- and intrachromosomal synteny detected via MCScanX (Wang et al. 2012). Minimum block size of 1 Mb, with each 1 Mb containing a majority of base pair in a syntenic block considered for linking.

Table 2. Results of individual contigs, not including unplaced scaffolds.

Contigs	
Contig	bp	Gaps	Ns	% N	
1	61,969,647	22,538	2,406,911	3.88	
2	58,466,665	19,556	2,017,778	3.45	
3	41,908,749	15,456	1,626,681	3.88	
4	33,349,723	11,535	1,358,494	4.07	
5	26,039,851	9429	1,210,825	4.65	
6	24,737,718	8728	1,058,287	4.28	
Cp	154,223	0	0	0.00	
Mt	255,307	6	600	0.24	

Integrity of draft assembly strategy

The topography of the genome assembly with respect to the gene density and repeat density illustrates this strategy was likely effective in preserving a significant degree of the genome structure of P. fendleri (Fig. 5). Furthermore, the sanity check that was performed by determining synteny between the unmasked target C. laxa assembly and the final P. fendleri assembly revealed that genome structure was not simply copied during super-scaffolding; indeed, for each P. fendleri pseudochromosome, synteny was observed with multiple C. laxa chromosomes (Fig. 6; Supplementary Table 8). Interestingly, synteny between this P. fendleri consensus with both A. thaliana and C. laxa revealed the possible presence of an inversion in several large regions. It is known that Brassicaceae underwent such processes regularly during evolution, particularly the compartmentalization of genomic rearrangements, which was observed in our draft consensus by several large regions of rearrangement when comparing synteny with A. thaliana (Murat et al. 2015). This trend is especially evident in diploid Brassicaceae, which often contain such compartmented genomic rearrangements after returning to a diploid state following an ancestral polyploidization (Cheng et al. 2012, 2014; Murat et al. 2015). Regardless, the characterization of this and other evolutionary phenomena would require further analyses for confirming the chromosomal localization of genomic features within the species, particularly optical mapping. The widescale outcrossing of P. fendleri allows it to serve as a rich model for studying structural genomics in Brassicaceae, as well as macro- and micro-scale feature comparisons between species. Furthermore, the characterization of P. fendleri grants a valuable perspective into the genomic characteristics that contribute to the potential of alternative oilseed crops.

Fig. 6. Interspecies synteny between P. fendleri and C. laxa (GenBank Acc.: GCA_024034495.1) and A. thaliana (Berardini et al. 2015). Detected via Satsuma2 (https://github.com/bioinfologics/satsuma2). Red links indicate synteny between P. fendleri and C. laxa, blue links indicate synteny between P. fendleri and A. thaliana. Scale of chromosome size is megabase pair (between layers b and c).

Supplementary Material

jkae114_Supplementary_Data

Acknowledgments

The authors thank the Pittsburgh Supercomputing Center that managed the Bridges-2 allocation used for this work, as well as the University of North Texas Genomics Center Core Facility for sequencing support.

Data availability

Corresponding data are available under NCBI BioProject PRJNA951248. The P. fendleri genome assembly was deposited in GenBank (NCBI Acc.: DAVZGE000000000). Generated paired-end RNA-seq data were deposited in the SRA: SRP430430. Both the chloroplast (Genbank Acc.: BK063463) and mitochondrial genomes (GenBank Acc.: OQ818196) are annotated and available on NCBI. All other third-party SRA data used are publicly available under the accessions listed in the text. Relevant program parameters and code are available in Supplementary Table 9.

Supplemental material available at G3 online.

Funding

This work was partially supported by Agricultural and Food Research Initiative (AFRI) grant no. 2021-67034-35137 to CRJ and 2021-67013-3777 to APA from the USDA National Institute of Food and Agriculture. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture. This work was computationally performed in part thanks to funding by National Science Foundation Award No. 1928147.
==== Refs
Literature cited

Alonge M , LebeigleL, KirscheM, JenikeK, OuS, AganezovS, WangX, LippmanZB, SchatzMC, SoykS. 2022. Automated assembly scaffolding using RagTag elevates a new tomato system for high-throughput genome editing. Genome Biol. 23 (1 ):258. doi:10.1186/S13059-022-02823-7.36522651
Alverson AJ , WeiX, RiceDW, SternDB, BarryK, PalmerJD. 2010. Insights into the evolution of mitochondrial genome size from complete sequences of Citrullus lanatus and Cucurbita pepo (Cucurbitaceae). Mol Biol Evol. 27 (6 ):1436–1448. doi:10.1093/MOLBEV/MSQ029.20118192
Anjani K . 2012. Castor genetic resources: a primary gene pool for exploitation. Ind Crops Prod. 35 (1 ):1–14. doi:10.1016/j.indcrop.2011.06.011.
Azeez A , ParchuriP, BatesPD. 2022. Suppression of Physaria fendleri SDP1 increased seed oil and hydroxy fatty acid content while maintaining oil biosynthesis through triacylglycerol remodeling. Front Plant Sci. 13 :1861. doi:10.3389/FPLS.2022.931310/BIBTEX.
Bankevich A , NurkS, AntipovD, GurevichAA, DvorkinM, KulikovAS, LesinVM, NikolenkoSI, PhamS, PrjibelskiAD, et al 2012. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J Comput Biol. 19 (5 ):455–477. doi:10.1089/CMB.2012.0021.22506599
Barclay A , GentryH, JonesQ. 1962. The search for new industrial crops II: Lesquerella (Cruciferae) as a source of new oilseeds on JSTOR. Economic Botany; [cited 2023 Mar 1]. Available from https://www.jstor.org/stable/4252334.
Bateman A , MartinMJ, OrchardS, MagraneM, AhmadS, AlpiE, Bowler-BarnettEH, BrittoR, Bye-A-JeeH, CukuraA, et al 2023. UniProt: the universal protein knowledgebase in 2023. Nucleic Acids Res. 51 (D1 ):D523–D531. doi:10.1093/NAR/GKAC1052.36408920
Berardini TZ , ReiserL, LiD, MezheritskyY, MullerR, StraitE, HualaE. 2015. The arabidopsis information resource: making and mining the “gold standard” annotated reference plant genome. Genesis. 53 (8 ):474–485. doi:10.1002/DVG.22877.26201819
Binder RG , ApplewhiteTH, KohlerGO, GoldblattLA. 1962. Chromatographie analysis of seed oils. Fatty acid composition of castor oil. J Am Oil Chemists’ Soc. 39 (12 ):513–517. doi:10.1007/BF02672540/METRICS.
Bryant DM , JohnsonK, DiTommasoT, TickleT, CougerMB, Payzin-DogruD, LeeTJ, LeighND, KuoTH, DavisFG, et al 2017. A tissue-mapped axolotl de novo transcriptome enables identification of limb regeneration factors. Cell Rep. 18 (3 ):762–776. doi:10.1016/J.CELREP.2016.12.063/ATTACHMENT/B24F06B8-7AC7-45AF-B605-0DA3B55AE55E/MMC11.ZIP.28099853
Bushmanova E , AntipovD, LapidusA, PrjibelskiAD. 2019. rnaSPAdes: a de novo transcriptome assembler and its application to RNA-seq data. GigaScience. 8 (9 ):giz100. doi:10.1093/GIGASCIENCE/GIZ100.31494669
Bushmanova E , AntipovD, LapidusA, SuvorovV, PrjibelskiAD. 2016. rnaQUAST: a quality assessment tool for de novo transcriptome assemblies. Bioinformatics. 32 (14 ):2210–2212. doi:10.1093/BIOINFORMATICS/BTW218.27153654
Camacho C , CoulourisG, AvagyanV, MaN, PapadopoulosJ, BealerK, MaddenTL. 2009. BLAST+: architecture and applications. BMC Bioinformatics. 10 (1 ):421. doi:10.1186/1471-2105-10-421.20003500
Campbell MS , HoltC, MooreB, YandellM. 2014a. Genome annotation and curation using MAKER and MAKER-P. Curr Protoc Bioinform. 48 (1 ):4.11.1–4.11.39. doi:10.1002/0471250953.BI0411S48.
Campbell MS , LawMY, HoltC, SteinJC, MogheGD, HufnagelDE, LeiJ, AchawanantakunR, JiaoD, LawrenceCJ, et al 2014b. MAKER-P: a tool kit for the rapid creation, management, and quality control of plant genome annotations. Plant Physiol. 164 (2 ):513–524. doi:10.1104/PP.113.230144.24306534
Cermak SC , BrandonKB, IsbellTA. 2006. Synthesis and physical properties of estolides from lesquerella and castor fatty acid esters. Ind Crops Prod. 23 (1 ):54–64. doi:10.1016/J.INDCROP.2005.04.001.
Cheng F , WuJ, FangL, WangX. 2012. Syntenic gene analysis between Brassica rapa and other Brassicaceae species. Front. Plant Sci. 3 .
Cheng F , WuJ, WangX. 2014. Genome triplication drove the diversification of Brassica plants. Hortic Res. 1 (1 ):14024. doi:10.1038/HORTRES.2014.24/41769057/41438_2014_ARTICLE_BFHORTRES201424.PDF.26504539
Cocuron J-C , AlonsoAP. 2023. 13C-labeling reveals non-conventional pathways providing carbon for hydroxy fatty acid synthesis in Physaria fendleri. J Exp Bot. 75 (6): 1754–1766. doi:10.1093/jxb/erad343.
Cocuron J-C , AndersonB, BoydA, AlonsoAP. 2014. Targeted metabolomics of Physaria fendleri, an industrial crop producing hydroxy fatty acids. Plant Cell Physiol. 55 (3 ):620–633. doi:10.1093/pcp/pcu011.24443498
Davison AG , BrittonMG, ForresterJA, DaviesRJ, HughesDTD. 1983. Asthma in merchant seamen and laboratory workers caused by allergy to castor beans: analysis of allergens. Clin Exp Allergy. 13 (6 ):553–561. doi:10.1111/J.1365-2222.1983.TB02637.X.
Dierig DA , TomasiPM, CoffeltTA, RayfordWE, LauverL. 2000. Registration of a Lesquerella fendleri germplasm with yellow seed coat color. Crop Sci. 40 (3 ):865–865. https://doi.org/null.
Dierig DA , WangGS, Crafts-BrandnerSJ. 2012. Dynamics of reproductive growth of lesquerella (Physaria fendleri) over different planting dates. Ind Crops Prod. 35 (1 ):146–153. doi:10.1016/J.INDCROP.2011.06.027.
Eilbeck K , MooreB, HoltC, YandellM. 2009. Quantitative measures for the management and comparison of annotated genomes. BMC Bioinformatics. 10 (1 ):67. doi:10.1186/1471-2105-10-67/FIGURES/8.19236712
Flynn JM , HubleyR, GoubertC, RosenJ, ClarkAG, FeschotteC, SmitAF. 2020. RepeatModeler2 for automated genomic discovery of transposable element families. Proc Natl Acad Sci U S A. 117 (17 ):9451–9457. doi:10.1073/PNAS.1921046117/SUPPL_FILE/PNAS.1921046117.SAPP.PDF.32300014
Grabherr MG , HaasBJ, YassourM, LevinJZ, ThompsonDA, AmitI, AdiconisX, FanL, RaychowdhuryR, ZengQ, et al 2011. Trinity: reconstructing a full-length transcriptome without a genome from RNA-seq data. Nat Biotechnol. 29 (7 ):644–652. doi:10.1038/NBT.1883.21572440
Grabherr MG , RussellP, MeyerM, MauceliE, AlföldiJ, di PalmaF, Lindblad-TohK. 2010. Genome-wide synteny through highly sensitive sequence alignment: Satsuma. Bioinformatics. 26 (9 ):1145–1151. doi:10.1093/BIOINFORMATICS/BTQ102.20208069
Gunstone FD . 2008. Disappearance. Lipid Technol. 20 (2 ):48. doi:10.1002/LITE.200800007.
Hoff KJ , StankeM. 2019. Predicting genes in single genomes with AUGUSTUS. Curr Protoc Bioinformatics. 65 (1 ):e57. doi:10.1002/CPBI.57.30466165
Horn PJ , LiuJ, CocuronJC, McGlewK, ThrowerNA, LarsonM, LuC, AlonsoAP, OhlroggeJ. 2016. Identification of multiple lipid genes with modifications in expression and sequence associated with the evolution of hydroxy fatty acid accumulation in Physaria fendleri. Plant J. 86 (4 ):322–348. doi:10.1111/tpj.13163.26991237
Huber T , FaulknerG, HugenholtzP. 2004. Bellerophon: a program to detect chimeric sequences in multiple sequence alignments. Bioinformatics. 20 (14 ):2317–2319. doi:10.1093/BIOINFORMATICS/BTH226.15073015
Jackman SD , VandervalkBP, MohamadiH, ChuJ, YeoS, HammondSA, JaheshG, KhanH, CoombeL, WarrenRL, et al 2017. ABySS 2.0: resource-efficient assembly of large genomes using a bloom filter. Genome Res. 27 (5 ):768–777. doi:10.1101/GR.214346.116.28232478
Jin JJ , YuW, YangJB, SongY, DepamphilisCW, YiTS, LiDZ. 2020. GetOrganelle: a fast and versatile toolkit for accurate de novo assembly of organelle genomes. Genome Biol. 21 (1 ):241. doi:10.1186/S13059-020-02154-5/FIGURES/6.32912315
Johnston C , García NavarreteLT, OrtizE, RomsdahlTB, GuzhaA, ChapmanKD, GrotewoldE, AlonsoAP. 2022. Effective mechanisms for improving seed oil production in pennycress (Thlaspi arvense L.) highlighted by integration of comparative metabolomics and transcriptomics. Front Plant Sci. 13 :943585. doi:10.3389/FPLS.2022.943585/FULL.35909773
Kiefer C , WillingEM, JiaoWB, SunH, PiednoëlM, HümannU, HartwigB, KochMA, SchneebergerK. 2019. Interspecies association mapping links reduced CG to TG substitution rates to the loss of gene-body methylation. Nat Plants. 5 (8 ):846–855. doi:10.1038/s41477-019-0486-9.31358959
Kim HU , ChenGQ. 2015. Identification of hydroxy fatty acid and triacylglycerol metabolism-related genes in lesquerella through seed transcriptome analysis. BMC Genomics. 16 (1 ):230. doi:10.1186/S12864-015-1413-8.25881190
Kim D , LangmeadB, SalzbergSL. 2015. HISAT: a fast spliced aligner with low memory requirements. Nat Methods. 12 (4 ):357–360. doi:10.1038/nmeth.3317.25751142
Korf I . 2004. Gene finding in novel genomes. BMC Bioinformatics. 5 (1 ):1–9. doi:10.1186/1471-2105-5-59/FIGURES/3.14706121
Kosugi S , HirakawaH, TabataS. 2015. GMcloser: closing gaps in assemblies accurately with a likelihood-based selection of contig or long-read alignments. Bioinformatics. 31 (23 ):3733–3741. doi:10.1093/BIOINFORMATICS/BTV465.26261222
Krzywinski M , ScheinJ, BirolI, ConnorsJ, GascoyneR, HorsmanD, JonesSJ, MarraMA. 2009. Circos: an information aesthetic for comparative genomics. Genome Res. 19 (9 ):1639–1645. doi:10.1101/GR.092759.109.19541911
Lee KR , ChenGQ, KimHU. 2015. Current progress towards the metabolic engineering of plant seed oil for hydroxy fatty acids production. Plant Cell Rep. 34 (4 ):603–615. doi:10.1007/S00299-015-1736-6/FIGURES/1.25577331
Li W , GodzikA. 2006. Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics. 22 (13 ):1658–1659. doi:10.1093/BIOINFORMATICS/BTL158.16731699
Luo R , LiuB, XieY, LiZ, HuangW, YuanJ, HeG, ChenY, PanQ, LiuY, et al 2015. Erratum to “SOAPdenovo2: an empirically improved memory-efficient short-read de novo assembler” [GigaScience, (2012), 1, 18]. GigaScience. 4 (1 ):1. doi:10.1186/S13742-015-0069-2/METRICS.25838885
Mandáková T , LiZ, BarkerMS, LysakMA. 2017. Diverse genome organization following 13 independent mesopolyploid events in Brassicaceae contrasts with convergent patterns of gene retention. Plant J. 91 (1 ):3–21. doi:10.1111/TPJ.13553.28370611
Manni M , BerkeleyMR, SeppeyM, SimãoFA, ZdobnovEM. 2021. BUSCO update: novel and streamlined workflows along with broader and deeper phylogenetic coverage for scoring of eukaryotic, prokaryotic, and viral genomes. Mol Biol Evol. 38 (10 ):4647–4654. doi:10.1093/MOLBEV/MSAB199.34320186
Mistry J , ChuguranskyS, WilliamsL, QureshiM, SalazarGA, SonnhammerELL, TosattoSCE, PaladinL, RajS, RichardsonLJ, et al 2021. Pfam: the protein families database in 2021. Nucleic Acids Res. 49 (D1 ):D412–D419. doi:10.1093/NAR/GKAA913.33125078
Mohamadi H , KhanH, BirolI. 2017. ntCard: a streaming algorithm for cardinality estimation in genomics data. Bioinformatics. 33 (9 ):1324–1330. doi:10.1093/BIOINFORMATICS/BTW832.28453674
Mokhtar MM , Abd-ElhalimHM, AllaliAE. 2023. A large-scale assessment of the quality of plant genome assemblies using the LTR assembly index. AoB PLANTS. 15 (3 ).
Murat F , LouisA, MaumusF, ArmeroA, CookeR, QuesnevilleH, CrolliusHR, SalseJ. 2015. Understanding Brassicaceae evolution through ancestral genome reconstruction. Genome Biol. 16 (1 ):262. doi:10.1186/S13059-015-0814-Y/FIGURES/4.26653025
Ohlrogge J , ThrowerN, MhaskeV, StymneS, BaxterM, YangW, LiuJ, ShawK, ShorroshB, ZhangM, et al 2018. Plant FA db: a resource for exploring hundreds of plant fatty acid structures synthesized by thousands of plants and their phylogenetic relationships. Plant J. 96 (6 ):1299–1308. doi:10.1111/tpj.14102.30242919
Olson ND , LundSP, ColmanRE, FosterJT, SahlJW, SchuppJM, KeimP, MorrowJB, SalitML, ZookJM. 2015. Best practices for evaluating single nucleotide variant calling methods for microbial genomics. Front Genet. 6 :235. doi:10.3389/FGENE.2015.00235/BIBTEX.26217378
Qin M , WuS, LiA, ZhaoF, FengH, DingL, RuanJ. 2019. LRScaf: improving draft genomes using long noisy reads. BMC Genomics. 20 (1 ):955. doi:10.1186/S12864-019-6337-2/TABLES/4.31818249
Reese J , ChristensonMK, LengN, SahaS, CantarelB, LindebergM, TamborindeguyC, MacCarthyJ, WeaverD, TreaseAJ, et al 2014. Characterization of the Asian citrus psyllid transcriptome. J Genomics. 2 (11 ):54–58. doi:10.7150/JGEN.7692.24511328
Reuscher S , FurutaT, Bessho-UeharaK, CosiM, JenaKK, ToyodaA, FujiyamaA, KurataN, AshikariM. 2018. Assembling the genome of the African wild rice Oryza longistaminata by exploiting synteny in closely related Oryza species. Commun Biol. 1 (1 ):162. doi:10.1038/s42003-018-0171-y.30320230
Roach MJ , SchmidtSA, BornemanAR. 2018. Purge Haplotigs: allelic contig reassignment for third-gen diploid genome assemblies. BMC Bioinformatics. 19 (1 ):460. doi:10.1186/S12859-018-2485-7/FIGURES/5.30497373
Robertson G , ScheinJ, ChiuR, CorbettR, FieldM, JackmanSD, MungallK, LeeS, OkadaHM, QianJQ, et al 2010. De novo assembly and analysis of RNA-seq data. Nat Methods. 7 (11 ):909–912. doi:10.1038/NMETH.1517.20935650
Román-Figueroa C , CeaM, PanequeM, GonzálezME. 2020. Oil content and fatty acid composition in castor bean naturalized accessions under Mediterranean conditions in Chile. Agronomy. 10 (8 ):1145. doi:10.3390/agronomy10081145.
Salywon AM , DierigDA, RebmanJP, de RodríguezDJ. 2005. Evaluation of new Lesquerella and Physaria (Brassicaceae) oilseed germplasm. Am J Bot. 92 (1 ):53–62. doi:10.3732/ajb.92.1.53.21652384
Sohn J , NamJW. 2018. The present and future of de novo whole-genome assembly. Brief Bioinformatics. 19 (1 ):23–40. doi:10.1093/BIB/BBW096.27742661
Song L , ShankarDS, FloreaL. 2016. Rascaf: improving genome assembly with RNA sequencing data. Plant Genome. 9 (3 ):plantgenome2016.03.0027. doi:10.3835/PLANTGENOME2016.03.0027.
Sork VL , Fitz-GibbonST, PuiuD, CrepeauM, GuggerPF, ShermanR, StevensK, LangleyCH, PellegriniM, SalzbergSL. 2016. First draft assembly and annotation of the genome of a California endemic oak Quercus lobata Née (Fagaceae). G3 (Bethesda). 6 (11 ):3485–3495. doi:10.1534/G3.116.030411.27621377
Stanke M , KellerO, GunduzI, HayesA, WaackS, MorgensternB. 2006. AUGUSTUS: ab initio prediction of alternative transcripts. Nucleic Acids Res. 34 (suppl_2 ):W435–W439. doi:10.1093/NAR/GKL200.16845043
Tamazian G , DobryninP, KrasheninnikovaK, KomissarovA, KoepfliKP, O’BrienSJ. 2016. Chromosomer: a reference-based genome arrangement tool for producing draft chromosome sequences. GigaScience. 5 (1 ):38. doi:10.1186/S13742-016-0141-6/2737417.27549770
Tang L , LiM, WuFX, PanY, WangJ. 2020. MAC: merging assemblies by using adjacency algebraic model and classification. Front Genet. 10 :506828. doi:10.3389/FGENE.2019.01396/BIBTEX.
Tang S , LomsadzeA, BorodovskyM. 2015. Identification of protein coding regions in RNA transcripts. Nucleic Acids Res. 43 (12 ):e78–e78. doi:10.1093/NAR/GKV227.25870408
Thorpe SC , KemenyDM, PanzaniRC, McGurlB, LordM. 1988. Allergy to castor bean: II. Identification of the major allergens in castor bean seeds. J Allergy ClinI mmunol. 82 (1 ):67–72. doi:10.1016/0091-6749(88)90053-X.
Tillich M , LehwarkP, PellizzerT, Ulbricht-JonesES, FischerA, BockR, GreinerS. 2017. Geseq—versatile and accurate annotation of organelle genomes. Nucleic Acids Res. 45 (W1 ):W6–W11. doi:10.1093/NAR/GKX391.28486635
Wang Y , TangH, DebarryJD, TanX, LiJ, WangX, LeeTH, JinH, MarlerB, GuoH, et al 2012. MCScanx: a toolkit for detection and evolutionary analysis of gene synteny and collinearity. Nucleic Acids Res. 40 (7 ):e49. doi:10.1093/NAR/GKR1293.22217600
Xie Y , WuG, TangJ, LuoR, PattersonJ, LiuS, HuangW, HeG, GuS, LiS, et al 2014. SOAPdenovo-Trans: de novo transcriptome assembly with short RNA-seq reads. Bioinformatics. 30 (12 ):1660–1666. doi:10.1093/BIOINFORMATICS/BTU077.24532719
Xu M , GuoL, GuS, WangO, ZhangR, PetersBA, FanG, LiuX, XuX, DengL, et al 2020. TGS-GapCloser: a fast and accurate gap closer for large genomes with low coverage of error-prone long reads. GigaScience. 9 (9 ):giaa094. doi:10.1093/GIGASCIENCE/GIAA094.32893860
Xue W , LiJT, ZhuYP, HouGY, KongXF, KuangYY, SunXW. 2013. L_RNA_scaffolder: scaffolding genomes with transcripts. BMC Genomics. 14 (1 ):604. doi:10.1186/1471-2164-14-604/FIGURES/6.24010822
Yang Y , SmithSA. 2013. Optimizing de novo assembly of short-read RNA-seq data for phylogenomics. BMC Genomics. 14 (1 ):328. doi:10.1186/1471-2164-14-328/TABLES/1.23672450
Zhu BH , SongYN, XueW, XuGC, XiaoJ, SunMY, SunXW, LiJT. 2016. PEP-scaffolder: using (homologous) proteins to scaffold genomes. Bioinformatics. 32 (20 ):3193–3195. doi:10.1093/BIOINFORMATICS/BTW378.27334475
Zhu BH , XiaoJ, XueW, XuGC, SunMY, LiJT. 2018. P_RNA_scaffolder: a fast and accurate genome scaffolder using paired-end RNA-sequencing reads. BMC Genomics. 19 (1 ):175. doi:10.1186/S12864-018-4567-3/FIGURES/7.29499650
Zimin Av , MarçaisG, PuiuD, RobertsM, SalzbergSL, YorkeJA. 2013. The MaSuRCA genome assembler. Bioinformatics. 29 (21 ):2669–2677. doi:10.1093/BIOINFORMATICS/BTT476.23990416
