Beyond the developmental predictions, we improved the model's precision and robustness with: 1) improved splicing outlier ground truth, 2) richer set of predictive features from SpliceAI/Pangolin, 3) replacing binary splice site usage with continuous usage features.

We predicted thousands of variants with a developmental impact genome-wide. In Genomics England, we identified a new candidate in FGFR1 for an individual with Kallmann syndrome. The variant’s effect is predicted to occur solely in an early developmental window that aligns with the disease mechanism.

Within the rare-disease cohort Solve-RD, high-impact brain-specific predictions are significantly enriched in individuals affected by neurodevelopmental disorders (NDD) in NDD-linked genes. The predicted developmental timing of the splicing disruption correlates with the clinical age of onset.

Our aberrant splicing prediction model AbSplice originally used adult data, but splicing is developmentally dynamic. We extended the framework to capture a class of variants impacting exclusively early stages - these have previously stayed off our radar.