NSMAP

NSMAP infers and quantifies expressed transcript isoforms directly from RNA-Seq data without requiring annotated reference genomes to enable analysis of transcriptomic complexity despite incomplete references.

Key Features:

• Isoform Structure Prediction: Identifies structures of expressed isoforms directly from RNA-Seq data, enabling discovery of novel splice variants.
• Simultaneous Expression Estimation: Estimates expression levels concurrently with isoform structure prediction to provide joint quantification and structure inference.
• Annotation-Free Analysis: Operates without annotated reference genomes to discover and quantify novel isoforms missed by annotation-dependent methods.
• Sparsity-Regularized Estimation: Integrates a sparsity term into expression level estimation to promote sparse isoform models.
• NSMAP Model: Implements a Nonnegativity and Sparsity constrained Maximum APosteriori (NSMAP) model for inference.
• Model Parameterization and Validation: Model parameters were estimated using real RNA-Seq datasets and validated in simulations reporting over 77% accuracy in identifying and quantifying expressed isoforms.

Scientific Applications:

• Alternative Splicing Studies: Identifies and quantifies different splice variants to study alternative splicing mechanisms.
• Gene Expression Regulation: Provides isoform-level expression profiles to investigate gene expression regulation and isoform diversity.
• Disease Research (MDS): Has been used to identify known and novel differentially expressed isoforms in studies of myelodysplastic syndromes (MDS).

Methodology:

NSMAP uses a Nonnegativity and Sparsity constrained Maximum APosteriori model with an integrated sparsity term, parameterized on real RNA-Seq datasets and evaluated by simulations reporting >77% accuracy in isoform identification and quantification.