Normal-Tissue Absence, Rather Than Cross-Patient Prevalence, Limits a Fixed Splice-Junction Panel in Hepatocellular Carcinoma: From a Pre-Specified In Silico Screen to a Genome-Wide Analysis

Background/Objectives: A fixed splice-junction panel for hepatocellular carcinoma (HCC) requires junctions recurring across patients and absent from normal tissue; absence is inferred from non-detection in finite references. Methods: Across 300 secreted, membrane-protein and HCC-associated genes, 1,263,305 junctions were screened for recurrence (≥30%) in 374 TCGA-LIHC tumours and absence in 50 matched normals and 251 GTEx liver samples. Results: No junction met the pre-specified criteria. Widely recurring junctions occurred in normal tissue; of 981 canonical junctions meeting both absence criteria, 740 were seen in one tumour. Post hoc, 8 of the 981 exceeded 11.63%, the prevalence below which non-detection in 50 normals does not demonstrate absence. None reached 30%; they reflect TERT expression, UGT1A first-exon choice, and a UGT2B readthrough of unresolved origin also present in normal breast. Twenty such junctions covered 63.9% of held-out tumours (post hoc), with absence assumed for most. Positive controls from other cancers were recovered. Genome-wide (post hoc), 18 junctions met all criteria, 17 at loci silent in normal liver. Conclusions: No panel junction was both recurrent and demonstrably absent from normal liver; genome-wide, such junctions mostly reflected expression of these silent loci, not tumour-specific splicing. Reporting the prevalence above which absence is demonstrable is recommended.