TxNova: recovery of recurrent unannotated intergenic splice loci from existing bulk RNA-seq alignments
Background. Reference catalogs such as GENCODE capture most stably expressed mammalian genes but may not include condition-restricted or low-abundance transcripts. Reads supporting unannotated intergenic splice junctions are generally absent from annotation-restricted gene-count matrices; transcript assemblers may reconstruct a subset as novel models, but those models are typically handled…
TxNova is a tool that extracts unannotated intergenic splice junctions from bulk RNA-seq alignments, which are often missing from reference catalogs like GENCODE. These junctions are not typically reconstructed by transcript assemblers and are usually handled separately from annotated gene-count data. TxNova works directly on existing BAM files, without the need for an external assembler, gffcompare, or workflow manager, making it a lightweight and efficient option for researchers.
To identify potential splice junctions, TxNova retains CIGAR N junctions from STAR/HISAT2 BAM files that appear in at least two samples but are absent from a comprehensive annotation. These junctions are then clustered into residual loci and counted alongside annotated genes in a unified matrix. Structure gates, such as canonical splice motifs, same-strand distance, coverage valley, bridging-junction absence, and minimum length, are used to filter out likely artifacts and yield structure-pass models.
An optional contrast filter can be applied to retain loci detected in treatment conditions but nearly silent in control conditions. Across four published mouse treatment arms, TxNova generated 464, 657, 789, and 594 residual splice loci. In one of these datasets (GSE221720), 45% of the loci (209/464) shared an exact intron with another series, while excluded junctions only matched 0.074% for intron placement.
For genes that had a leak junction, TxNova achieved 87.6% sensitivity (176/201) and 99.4% specificity (176/177). These values represent sensitivity and precision estimates, respectively.
The recovered models are partial reconstructions that require further validation through cloning, RACE, or targeted proteomics. Overall, TxNova offers a reproducible method for generating intergenic residual-locus catalogs from existing BAM files, with four downloadable mouse injury/infection catalogs. However, while cross-series recurrence and coordinate-based null controls support the tool's reproducibility, they do not by themselves establish biological validity.
The optional contrast step provides a useful screen for candidate leads that warrant experimental validation.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
