De novo transcriptome and annotation files for Colobanthus apetalus

This transcriptome was assembled from 12 Illumina NovaSeq 6000 libraries generated as 2 × 150 bp paired-end reads. These libraries were produced from RNA samples extracted from leaves of Colobanthus apetalus during a water restriction experiment, with four samples each of control (100% field capacity), moderate stress (>70% field capacity), and severe stress (>50% field capacity) conditions.

Raw reads were quality-filtered with fastp v1.0.1 to remove low-quality bases, excessive Ns, adapter contamination, and poly(A) tails. Filtered reads were then assembled independently using Trinity v2.15.2, SOAPdenovo v1.0.4, Trans-ABySS v2.0.1, RNA-Bloom v2.0.1, and rnaSPAdes v4.2.0, all with default parameters.

The resulting assemblies were concatenated and subjected to successive redundancy-reduction steps, including exact sequence deduplication, transcript length filtering (minimum 300 bp), high-stringency clustering with CD-HIT v4.8.1 at 99% sequence identity, containment removal, and expression-based filtering using Salmon v1.10.3. Only transcripts meeting the minimum expression threshold were retained.

Candidate open reading frames (ORFs) were identified with TransDecoder 2 (github.com/Markusjsommer/TD2; Mao et al., 2025 - https://www.biorxiv.org/content/10.1101/2025.04.13.648579v1), and the longest predicted peptides were searched against a custom Viridiplantae-filtered UniRef100 database using MMseqs2 (commit befcb1130a73ddba247fe3eaa504e7ffb790d150). This database was generated by downloading UniRef100 and its associated taxonomy information through MMseqs2 and filtering entries assigned to Viridiplantae. These homology results were used as supporting evidence to retain high-confidence ORFs in TransDecoder 2.

To further reduce transcriptome redundancy, transcripts were searched against the proteome of closely related species Silene latifolia using DIAMOND blastx v2.0.15, and the top two matching ORFs were retained. ORFs without matches to the sister-species proteome were retained as putatively novel only if they showed strong expression support (≥100 total mapped reads across all samples) and encoded peptides of at least 150 amino acids.

The transcripts and predicted proteins were further annotated with hmmscan (HMMER v3.4) against the Pfam-A database (release 13-01-2026), and the Mercator v4.8 database (Schwacke et al., 2019 - https://doi.org/10.1016/j.molp.2019.01.003).

Additional fasta and text manipulation steps, including duplicate removal and identifier standardization, were performed using SeqKit v2.3.0, Seqtk v1.4-r122, BBMap v39.33, and standard Bash command-line utilities.

The script with commands to generate the files in this dataset are in 10.5281/zenodo.18874676.