Blueprint for Phasing and Assembling the Genomes of Heterozygous Polyploids: Application to the Octoploid Genome of Strawberry

Key results

1. ShownFaRR1 is a synthetic haploid reference built by taking the more contiguous phase-1 or phase-2 pseudomolecule for each of the 28 chromosomes. It places nearly all sequence in chromosome-scale pseudomolecules, with few gaps.

   786.5 Mb in total, with 99.5% in 28 chromosome-scale pseudomolecules and 21 unanchored contigs. Seven pseudomolecules (25%) are single gapless sequences; mean 2.5 contigs per pseudomolecule after gap-filling. LR_Gapcloser with HiFi reads closed 17 of 82 (phase-1) and 20 of 81 (phase-2) gaps. Phase assembly N50 was 11 Mb before and 23 Mb after scaffolding and gap-filling. · Introduction; Assembly Scaffolding and Polishing

2. ShownParental Illumina k-mers split the HiFi reads by parent, with reads lacking a parental signal going to both bins. Each bin was then assembled separately with FALCON, giving two haplotype assemblies of near-identical size and contiguity.

   31% of HiFi reads were classed as parent-specific (at least 100 parental 21-mers, at least 80% from one parent); the bins held 70.8 and 69.8 Gb, about 80X haploid coverage each. After chimera splitting and redundancy removal: phase-1 136 contigs, 784.7 Mb, N50 11.0 Mb; phase-2 131 contigs, 784.0 Mb, N50 10.7 Mb. ALLMAPS scaffolded 99.5% of each into 28 pseudomolecules. · HiFi Trio-Binning; Phased Octoploid Genome Assembly; Assembly Scaffolding and Polishing; Trio Binning

3. ShownStrict overlap settings (99.5% identity for detection, 99.8% for filtering) limited subgenome chimeras but did not prevent them, and left redundant artifact contigs. A 3.6 M-variant recombination map anchored to the contigs located the chimeric junctions and binned contigs by chromosome.

   Initial FALCON output: 979 Mb in 310 contigs (phase-1) and 948 Mb in 272 (phase-2), against an approximate 800 Mb haploid size; over half of contigs lay fully within larger ones at >99% but <99.8% identity. ALLMAPS flagged 5 (phase-1) and 4 (phase-2) chimeras, each spanning two linkage-group regions; 99.1% and 99.2% of sequence was then binned into chromosome pools. · Phased Octoploid Genome Assembly; Recombination Map Anchoring; Subgenome Chimeras and Homologous Allelic…

4. ShownFaRR1 matches the diploid F. vesca v4.0 reference on BUSCO completeness, and the phased subgenome assemblies score higher than Camarosa v1 and F. vesca Hawaii 4 on the LTR Assembly Index. FaRR1 also fixes scaffolding and allelic-contig errors seen in Camarosa v1.

   BUSCO core eudicot set: 98.1% complete for both FaRR1 and F. vesca v4.0. LAI 18.2 to 19.33 for every subgenome of both phases (comparator values are not given in the text; see Table 3 / Supplementary Table 4). Plotted against the recombination map, FaRR1 resolves major scaffolding errors on 2C, 2D, 4A and 6C, and interleaved allelic contigs on 7A and 7C, both seen in Camarosa v1. · Assembly Quality and Completeness; Assembly Quality Assessment

5. SuggestedRe-annotation with IsoSeq-prioritised transcript evidence and a custom EDTA repeat library gives about 102k gene models per haploid genome. The authors report that these models agree better with the evidence than the Camarosa v1 models do.

   Repeats: 38.42% of the combined phase assemblies, mostly LTR elements. Of 248,061 MAKER models, 203,170 remained after removing TE-related and fragmentary ones (101,721 phase-1, 101,449 phase-2, 101,793 FaRR1), about 94% of the Camarosa v1 count. The quality gain is shown only as cumulative AED curves, with no summary values. · Genome Annotation; Transcriptome Assembly; Gene Annotation