domains / families / science / outlier-splicing-detection

outlier-splicing-detection

separate

Detects aberrant splicing in single rare-disease patients vs a control panel using FRASER 2.0 (Bioconductor; Beta-binomial autoencoder on Intron Jaccard Index, default delta cutoff 0.1, q hyperparameter), OUTRIDER (gene-level outlier expression via autoencoder denoising), LeafcutterMD (Dirichlet-multinomial outlier mode of LeafCutter for annotation-free junctions), and DROP (Snakemake pipeline integrating FRASER2 + OUTRIDER + monoallelic expression for clinical diagnostics). The statistical model is fundamentally different from differential splicing — single-sample-vs-cohort outlier detection rather than two-group comparison. Standard tool in EU rare-disease (Solve-RD) and NIH UDN programs. Use when applying RNA-seq to undiagnosed Mendelian disease, validating predicted splice variants in clinical samples, or detecting cryptic splicing in disease tissue.

When to use

• Single patient or small case series vs control panel
• Question: "What is aberrant in this patient?"
• Statistical model: single-sample p-value vs cohort distribution
• Two well-defined groups, n>=3 each
• Question: "What differs between groups?"
• Statistical model: two-group LRT or related

When not to use

• two-group differential splicing
• cohort <20 samples
• solid tumor splicing biomarker with n>=10 vs cohort

Skill page

Source