Reproducible cross-aligner detection of the FMR1-217 junction across human cell contexts in public RNA-seq data

Public RNA-seq repositories enable large-scale reuse, but cross-study detection of splice-junction evidence is often undermined by uneven sequencing depth, inconsistent quality control, and the concern that junction calls reflect alignment artifacts. We present a reproducible, quality-controlled framework to detect the FMR1-217 mis-splicing event across a harmonized human RNA-seq compendium of 1,181 quality-controlled runs from 93 ENA BioProjects and one in-house fragile-X premutation cell line. FMR1-217 is a CGG expansion-linked FMR1 transcript in which exon 1 is joined to a pseudo-exon within intron 1; it is of therapeutic interest because it can be suppressed by antisense oligonucleotides, with restoration of full-length FMR1 and FMRP. Detection was defined by a fixed splice-junction rule after a housekeeping-depth filter, which strongly enriched for detectable signal (19.8% of QC-pass versus 2.3% of QC-fail runs). To test robustness to alignment choice, we re-quantified FMR1-217 in 814 runs using an independent aligner (HISAT2); detection was almost perfectly concordant with STAR (97.1%, Cohen’s κ = 0.91), establishing that the signal is not an artifact of a single pipeline. Under identical rules, FMR1-217 detection varied across cell contexts and was highest in iPSC, although this variation was partly confounded with study of origin; a significant association with fragile X syndrome was observed in blood, the one stratum permitting an unconfounded within-study comparison, but not in the pooled cohort. A locus-restricted multi-omics summary of the premutation line C13589, which was junction-negative and showed low FMR1 methylation consistent with its smaller repeat, anchors the RNA-level findings to local epigenomic and structural context. This framework provides a defensible basis for benchmarking FMR1-217 detection across studies and for prioritizing datasets and model systems for targeted follow-up, including as a potential pharmacodynamic readout.