Abstract
Purpose :
Mutations in the introns of IRD genes are causing disease in ~20% of all IRD cases. There are roughly 1.5 mln rare deep intronic variants (RDIVs) in the genome. Prediction of RDIVs functional effect is still very uncertain. One of the most frequently used algorithms, the SpliceAI, in our experience predicted only 1 out of 18 experimentally confirmed pathogenic RDIVs, indicating a high false negative prediction rate and the need for improvement. We aim to combine high throughput experimental approach with computational approaches to significantly enhance prediction of functional effect of RDIVs.
Methods :
We used a previously described high throughput splicing assay (HTSA) of stable cell line transfection and compared results with 5 most precise in silico algorithms: NNSplice, SpliceAI, PDIVAS, Pangolin and Alamut. HTSA allows to evaluate ~1000 variants simultaneously. To increase robustness of the method, we optimized the technique using short-term transient transfection and investigation of minigene transcripts which allows for interrogation of 10,000s RDIVs at once.
Results :
Intronic variants were chosen from known described cases as positive controls and from patients with recessive IRDs with 1 confirmed pathogenic mutation in exon of the IRD gene. Following RDIVs showed cryptic exon (CE) inclusion after several different methods of interrogation: ABCA4 Intron1 c.66+3186C>T (CE length 44 bp), c.67-2203G>A (CE 79 bp), Intron14 c.2160+998T>C (CE 162 bp), EYS Intron31 c.6424+73718_6424+73724del (CE 82 bp), USH2A Intron64 c.14133+8635C>A (CE 80 bp), VPS13B Intron21 c.3083-14717C>T (CE 75 bp). All predicted variants led to aberrant splicing. Transient and stable transfection showed comparable results: average Pearson correlation was r=0.94 in two replicates, p<0.0001
Conclusions :
Overcoming the bottleneck of number of variants checked at one time we can quickly get results of bulk RDIVs interpretation: 4000 in RPE65 and 8000 in both RPGR and NMNAT1 genes which is useful for clinicians and bioinformaticians in diagnosis confirmation for inclusion patients in ongoing interventional clinical trials. We anticipate the cross-referencing experimental data with splicing prediction algorithms will lead to the improvement of the RDIVs pathogenicity predictions and to a more accurate diagnostics of IRDs and other genetic mendelian disorders.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.