*3.3. Mitochondrial Genome Assembly and Assembly Check*

The initial assembly was performed by Unicycler 0.4.8 [13] with the default parameters. After assembly, we estimated the mean read coverages of all contigs by mapping all reads to contigs with minimap 2.17 [46] with the option "asm20", recommended by the author of minimap for mapping PacBio CCS reads. The list of coverages for all contigs was created based on the mapping, by the command "DepthOfCoverage" from GATK 3.8 [47]. Potential mitochondrial contigs were identified based on a similarity search using BLASTN 2.9.0 [48] (e-value threshold of 10−3, word size 7 bp). The following genomes were used as query: The mitochondrial genomes of *Fallopia multiflora*, the only publicly available assembled mitochondrial genome from the family Polygonaceae, *Nepenthes* × *ventrata* (N. *ventricosa* × *Nepenthes alata*) and *Arabidopsis thaliana* and the plastid genome of *F. esculentum*. The latter is necessary in order to distinguish between plastid and mitochondrial contigs and to find regions of plastid origin in the mitochondrial genome. The check of the assembly was performed using dedicated genome assembler, which we named Elloreas (abbreviation for ELongating LOng REad ASsembler). It is deposited at https://github.com/shelkmike/Elloreas. Elloreas was run with parameters "sequencing\_technology hifi\_pacbio", "minimum\_length\_of\_mapped\_read\_part 8000", "minimum\_read\_similarity 99%", "contig\_edge\_size\_to\_use\_for\_mapping 15000". As starters for Elloreas, we used several random mitochondrial (based on the similarity search described above) contigs from assemblies created by Canu 2.0 [49] and Falcon 1.1.0 [50]. The assembly of Canu was performed with all reads, the parameter "genomesize" was set to 1.5 Gbp, which is the approximate nuclear genome size of *F. esculentum* [51]. The assembly by Falcon was done using reads with a median k-mer coverage of at least 50. The estimated genome size for Falcon was initially set to 1.5 Gbp, but the sum of produced contigs lengths was only 400 Mbp, due to the removal of many nuclear reads by Stacks. Therefore, we reran the Falcon, setting the estimate of the genome size to 400 Mbp. Other parameters of Falcon were set to the values recommended by the authors of Falcon for CCS reads.
