<i>Jinshaia</i><i> </i><i>abbreviata</i><b><i> </i></b>genome and annotation files

This dataset contains the complete genome annotation of *Jinshaia abbreviata*, a species of loach (Actinopterygii), providing a foundational genomic resource for evolutionary, ecological, and functional studies in loaches and related fish species.

### 1. Genome assembly
The *J. abbreviata* genome assembly (total length: 593.69 Mb) is provided in `Jinshaia_abbreviata.genome.fasta.gz`.

### 2. Repetitive element annotation
Repetitive sequences were identified using an integrated strategy combining homology-based searches against curated databases with de novo repeat prediction. Homology detection was performed using RepeatMasker v4.1.5 with the RepBase library, complemented by de novo repeat library construction implemented in RepeatModeler v2.0.4. The two sets of predictions were merged to generate a non-redundant annotation.

This comprehensive analysis revealed that transposable elements (TEs) occupy 181.65 Mb of the *J. abbreviata* genome, accounting for 30.60% of the total assembly length. Among the classified TEs, DNA transposons were the most abundant (13.20%), followed by LTR retrotransposons (8.29%), LINE retrotransposons (3.98%), and SINE retrotransposons (1.73%). Additionally, satellite repeats constituted 1.90% of the genome, while approximately 3.11% of the assembled sequences were annotated as unclassified repeats.

### 3. Protein-coding gene annotation
Structural annotation was performed through a multi-evidence integration pipeline combining ab initio prediction, homology-based inference, and transcriptomic evidence. Ab initio gene models were generated using Augustus v3.3.2 and Genscan. For homology-based evidence, protein sequences from three closely related loach species with high-quality, widely cited annotations (*Misgurnus anguillicaudatus*, *Triplophysa dalaica*, and *T. rosa*) were aligned to the genome using Exonerate v2.2.0 and Liftoff v1.6.3. Transcriptomic evidence was derived from multi-tissue RNA-seq data, which were aligned using HISAT2 v2.1.0, assembled into transcripts with StringTie v1.3.5, and finally polished with PASA v2.3.2 for consensus isoform assembly. These heterogeneous evidence tracks were integrated and reconciled via the MAKER2 v2.31.10 pipeline, which trained its internal predictors and produced a preliminary consensus gene set. Final high-confidence model selection and refinement were performed using the HiFAP tool.

This integrated approach yielded a final annotated set of **23,167 high-confidence protein-coding genes**. The gene structures have an average length of 12,411 bp, containing an average of 10.52 exons per gene. A key accuracy metric is the mean exon length (294.79 bp), which shows strong agreement with the length distribution derived solely from transcript evidence (322.23 bp). To quantify annotation confidence, we analyzed the evidence support for each gene: 73.8% (17,102 genes) had over 50% of their coding sequence concurrently supported by all three evidence types (ab initio, homology, and transcriptomic), marking them as highest-confidence models. Only 212 genes (0.92%) relied on a single type of low-overlap ab initio prediction. Benchmarking with BUSCO (actinopterygii_odb10) confirmed high functional completeness, with **96.2% of conserved actinopterygian orthologs successfully identified**.

The protein sequences and coding sequences (CDS) are provided in:
- `Jinshaia_abbreviata.protein.fasta.gz` (or `Jinshaia_abbreviata.pep.fasta.gz`)
- `Jinshaia_abbreviata.cds.fasta.gz`

### 4. Non-coding RNA annotation
Non-coding RNAs (ncRNAs) were systematically identified using tRNAscan-SE and Infernal cmscan tools with reference to the Rfam database. A total of **147,979 ncRNA genes** were predicted, whose combined length accounts for approximately 2.81% of the genome. Among these, transfer RNAs (tRNAs) were the most abundant in copy number (22,817), while ribosomal RNAs (rRNAs) contributed the greatest total length (13.9 Mb, 2.35% of the genome), driven predominantly by a high copy number of 5S rRNA genes. Furthermore, we annotated 3,994 small nuclear RNAs (snRNAs) and 3,510 microRNA precursors. This catalog offers a comprehensive resource for studying transcriptional and post-transcriptional regulatory mechanisms in *J. abbreviata*.

### 5. Complete annotation file
All structural annotations (repeats, genes, ncRNAs) are merged into a single GFF3 file:
- `Jinshaia_abbreviata.gff3.gz`

### 6. Chromosomal distribution
Chromosomal distributions of gene density, GC content, and repeat sequence density were plotted across all 25 *J. abbreviata* chromosomes (provided in Fig. 3 of the associated manuscript).