Oryziinae

The six bacterial isolates used in the real-read tests each belong to a different species: Acinetobacter baumannii, Citrobacter koseri, Enterobacter kobei, an unnamed Haemophilus species (given the placeholder name Haemophilus sp002998595 in GTDB R202 [28 , 29 ]), Klebsiella oxytoca and Klebsiella variicola. Sequencing was previously described in Wick et al. (2021) [19 ]. Briefly, isolates were cultured overnight at 37°C in Luria-Bertani broth and DNA was extracted using GenFind v3 according to the manufacturer’s instructions (Beckman Coulter). The same DNA extract was used to sequence each isolate using three different approaches: ONT ligation, ONT rapid and Illumina (S11(A) Fig). For ONT ligation, we followed the protocol for the SQK-LSK109 ligation sequencing kit and EXP-NBD104 native barcoding expansion (Oxford Nanopore Technologies). For ONT rapid, we followed the protocol for the SQK-RBK004 rapid barcoding kit (Oxford Nanopore Technologies). All ONT libraries were sequenced on MinION R9.4.1 flow cells. ONT read sets were basecalled and demultiplexed with Guppy v5.0.7, using the super-accuracy model. For Illumina, we followed a modified Illumina DNA Prep protocol (catalogue number 20018705), whereby the reaction volumes were quartered to conserve reagents. Illumina libraries were sequenced on the NovaSeq 6000 using SP reagent kits v1.0 (300 cycles, Illumina Inc.), producing 150 bp paired-end reads with a mean insert size of 331 bp. The resulting Illumina read pairs were shuffled and evenly split into two separate read sets, which were combined with the ONT read sets to produce two independent hybrid read sets (S11(B) Fig). We repeated this process (from culture to sequencing) to generate another two hybrid read sets for a total of four hybrid read sets per isolate. All reads are available in the manuscript’s data repository (bridges.monash.edu/articles/dataset/Polypolish_paper_dataset/16727680).
For each hybrid read set, we performed a long-read-only assembly using Trycycler v0.5.0 [3 ] and Medaka v1.4.3 [8 ], following the instructions in Trycycler’s documentation (S11(C) Fig and S6 Table). One of the ONT read sets for K. oxytoca MSB1_2C had very low depth (10×) and was therefore not able to yield a high-quality long-read-only assembly, leaving only three assemblies for this genome. We were able to produce four complete (circularised) long-read-only assemblies for the other five genomes, giving a total of 23 assemblies which served as the ‘unpolished’ assemblies in our real-read tests.
Polished genome sequences were generated by running the short-read polishers as described above (S11(D) Fig). For the single-tool tests, each polisher was run consecutively three times on each assembly. For the greedy combination tests, each polisher (excluding the hybrid polishers and wtpoa which performed poorly in the single-tool tests) was run once on each genome, and the best-performing polisher was defined as the one with the smallest total pairwise distance in its output assemblies. The best-performing polisher’s output was then used as input for another round of polishing until there were no more improvements. The greedy combination tests were then performed again with Polypolish excluded.
To assess the quality of real-read assemblies, we used the edlib [26 ] library to perform a global alignment of the chromosome sequences for all pairwise combinations within each genome (S11(E) Fig). The total distance was used as a metric of assembly quality, with lower values being better and a value of zero indicating that all assemblies for the genome were identical. We also ran ALE [18 ] on each real-read assembly using short-read alignments from BWA-MEM [16 ].

Isolates that carried aerobactin were considered for long-read sequencing. To capture the diversity of the putative aerobactin-encoding plasmids, 16 isolates were selected for Oxford Nanopore sequencing based on the diversity in STs and AbSTs, 4 of which were from Thailand. DNA was extracted from pure cultures using the GenFind v3 kit (Beckman Coulter Life Sciences) on a Biomek i7 instrument, using the protocol ‘DNA extraction from bacteria using GenFind v3’. Library preparation was done using the SQK-LSK109 ligation sequencing kit. The sequencing was performed on an Oxford Nanopore GridION instrument (Oxford Nanopore Technologies), using a MinION R9.4.1 flow cell. Guppy version 5.0.14 (Oxford Nanopore Technologies) was used for basecalling and demultiplexing, using the super-accuracy basecalling model.
The genomes that were both Illumina and Nanopore sequenced were subjected to hybrid assembly. First, the long reads were quality controlled with NanoPlot [31 (link)] version 1.33.1. Then, Filtlong [32 ] version 0.2.0 was used to discard the lowest 10 % of reads based on length and quality. Unicycler was used to generate hybrid assemblies based on the filtered long reads and untrimmed Illumina reads. If the hybrid assembly failed, Filtlong was run again and set to remove the lowest 20 % of reads.
Genomes that were either incomplete after hybrid assembly or failed to assemble twice were subjected to long-read assembly and consensus analysis using Trycycler [33 (link)] version 0.5.1. Briefly, Trycycler was used to generate 12 subsets of reads, where each set of four subsets were independently assembled using Minipolish [34 (link)] version 0.1.2, Flye [35 (link)] version 2.9 and Raven (https://github.com/lbcb-sci/raven) version 1.6.1. Then, the contigs were clustered, and potential outliers were removed. If a cluster was represented by less than four contigs, a new subset of 24 read sets were generated and assembled as above. The contig clusters were then reconciled and aligned, before a consensus was made. The resulting long-read assembly was polished by using Medaka (Oxford Nanopore Technologies) version 1.4.4 and two rounds of Pilon [36 (link)] version 1.23.

Genomic DNA of the strains sequenced in this study was extracted and purified using Genomic-tips 100/G and a genomic DNA buffer set (Qiagen) from overnight cultures in lysogeny broth (LB) at 37 °C following the manufacturer’s protocol, with minor modifications including the addition of SDS (final concentration 1 %) after adding buffer B2 and incubation for 1 h at 50 °C. For short-read sequencing, libraries were prepared using the Nextera XT DNA sample prep kit (Illumina) or NEBNext Ultra II FS DNA library preparation kit (New England Biolabs) and sequenced using the Illumina MiSeq platform to generate paired-end sequence reads (301 bp ×2). Reads were trimmed by Platanus trim [27 (link)] with default parameters and assembled by Platanus_B_v1.3.1 [28 (link)]. Scaffolds ≥300 bp were used in this study.
To determine complete genome sequences, size selection of purified DNA was performed using magnetic beads (AMPure XP; Beckman Coulter) to obtain longer DNA fragments. Sequencing libraries were prepared using a rapid barcoding kit (SQK-RBK004), sequenced using the R9.4.1 flow cell with the Oxford Nanopore Technologies (ONT) MinION platform, and base-called using Guppy GPU ver. 3.4.5. (ONT). Reads were trimmed using NanoFilt [29 (link)] with the following parameters: minimum length=7 000 bp, minimum quality score=10 and 5'-terminal 100 bases cutting. For sequencing of three strains (93_161312, CEC13091 and F690), only≥15 kb reads were used for assembly to gain better results. For sequencing of strain F765, the minimum length was changed to 2000 bp to salvage its small plasmid sequence. ONT read assembly and polishing were performed using the microPIPE pipeline [30 (link)]. In brief, trimmed ONT reads were assembled using Flye (v2.8.3) [31 (link)] with the option ‘--plasmids’ and polished with ONT reads using four iterations of Racon (v1.4.20) [32 (link)] followed by one iteration of Medaka (v1.4.3) (GitHub – https://github.com/nanoporetech/medaka). Output contigs were further refined with Illumina short reads using NextPolish (v1.3.1) (GitHub – https://github.com/Nextomics/NextPolish). As the chromosome of strain 93_161312 and the plasmids of strain F690 were not circularized, manual curation was performed using Minimap2 [33 (link)], Integrative Genomics Viewer (igv) [34 (link)] and GenomeMatcher v3.0.2 [35 (link)] to obtain their circular chromosome and plasmid sequences.

Data quality control and assembly were performed using an established pipeline [28 (link)]. Briefly, raw reads were trimmed using Trimmomatic (version 0.39) with default parameters settings [29 (link)]. Assemblies were generated by unicycler (v0.3.0b) [30 (link)], and checked for standard quality parameters using quast (version 5.0.2) [31 (link)]. Genomes that met all the following quality criteria were included in the downstream analyses: estimated mean read coverage cut-off of ≥30×, 1.6 Mb ±160 kb genome assemblies, and contig count ≤200. The mean assembly length was 1 642 135 bp, with a mean read coverage of 98.2-fold, ranging from 30.2- to 269.4-fold. The mean G+C content was 30. 85 mol%, ranging from 30.06 to 31.13 mol%. A long-read assembly of strain N18-1277 was generated using Flye 2.8.1 [32 (link)], long-read polished with Medaka v1.5.0 (https://github.com/nanoporetech/medaka) and short-read polished with Polypolish v0.5.0 [33 (link)] and polca implemented in MaSuRCA v4.0.4 [34 (link)].