Sample preparation and genome sequencing

Body parts of single B. minax males were dissected to obtain DNA for whole-genome sequencing. Total DNA was extracted using the Magen HiPure Insect DNA Kits (D3129-02) (Guangzhou, China) according to the manufacturer’s instructions. Samples were sent to the GeneDenovo company (Guangzhou, China) (https://www.genedenovo.com/) for library construction and genome sequencing. Genome sequencing was carried out with the combination of Next-generation sequencing (NGS) using the Hiseq 2500 and third-generation sequencing adopting the Pacbio RSII. Briefly, five libraries were constructed, making two duplicated short fragment libraries (450 bp + 800 bp) and three long mate-pair libraries (2 kb + 5 kb + 10 kb), which produced in total 102G nucleotide bases (Supplementary file ¹, Table ^S1-1) and covered an estimated 300 × of the genome size (Supplementary file ¹, Figure ^S1-1). The genome size was estimated at about 331 Mb based on Kmer (k = 17) short fragmentary libraries analysis (Supplementary file ¹, Figure ^S1-2). For third-generation sequencing, two libraries were constructed, five SMRT cells were sequenced, total 5G raw data were obtained, and the genome coverage reached 15 × (Supplementary file ¹, Table ^S1-2); and SMRT analysis software (version 2.3.0) (https://www.pacb.com/products-and-services/analytical-software/smrt-analysis/) provided from Pacbio was used for the sequencing quality control. The genome assembly was divided into two steps: (1) Platanus (version1.2.1)³⁵ was used to assemble Illumina data and GapCloser (v1.10)³⁶ was utilized to extend the contig length; (2) PBjelly (PBSuite_15.8.24)³⁷ was used to extend the scaffolds and fill the gaps by combining the third-generation sequencing data. Totally, we obtained a genome size of 340 Mb, which is close to the estimated genome size. The numbers of contigs/scaffolds were 38,509/8019 and the contig/scaffold N50 reached 23 kb/1.6 Mb (Supplementary file ¹, Table ^S1-3). The genome sequences have been submitted to NCBI (SAYV00000000). The repetitive sequences account for 20.97% of the genome with using de novo prediction (RepeatModeler³⁸ and LTR-FINDER³⁹), RepBase⁴⁰-based homology prediction (RepeatMasker⁴¹ and RepeatProteinMask⁴²), and tandem repeats finder (TRF⁴³) (Supplementary file ¹, Table ^S1-4 and Table ^S1-5). Several methods of de novo prediction, homology-based gene prediction, and cDNA/EST prediction were used to predict gene structure after excluding the repetitive sequences (Supplementary file ¹, Figure ^S1-3). Softwares from three systems (Augustus 2.7⁴⁴, Genscan 1.0⁴⁵, and Glimmer HMM 3.0.1⁴⁶) were used for de novo prediction of gene models. MAKER 2.28 software⁴⁷ was applied to integrate all gene sets into a non-redundant and more complete one. Lastly, the BUSCO method⁴⁸ was used to estimate the reliable degree of gene models, and the complete single-copy BUSCO scores reached 98.5% in the assembled gene set and 94.9% in the annotated gene set, respectively (Supplementary file ¹, Table ^S1-6). After all these analyses, a set of 12,533 gene models was obtained and used for identification of chemosensory genes.