Vision-Based Genomic Model for Copy Number Variant Pathogenicity Prediction.

Spatial representation of CNVs as base-pair-resolution multi-track images → vision model (TESSERACT) exploiting co-occurrence patterns across genomic annotation tracks

Current CNV pathogenicity prediction approaches reduce variants to regional numerical features, losing the positional structure exploited by clinical experts. TESSERACT represents each CNV as a base-pair-resolution multi-track image and applies a vision model to capture spatial patterns across annotation tracks. Evaluated on clinically interpreted CNVs, the model outperforms existing numerical approaches and better reproduces expert decisions.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The image-based CNV representation approach is conceptually elegant and biologically motivated — it mimics the way an expert geneticist visually examines a CNV on a genomic browser. Preprint to follow for clinical validation and peer-reviewed publication, but the approach is promising for improving automated CNV classification in diagnostic pipelines.

Clinical impact: 2/3 · Evidence strength: 2/3 · Novelty: 2/2 · Sample size: 1/1 · Publication status: -1/1 → Total: 6/10