What is a biomarker?
A biomarker is a measurable biological characteristic that can help describe disease state, progression, or response to treatment. In SMA research, biomarkers can help track molecular changes and therapeutic effects.
GAYAR
Menu
Area of Focus
Spinal Muscular Atrophy (SMA)
GAYAR's work in SMA focuses on molecular and computational questions around biomarker discovery, SMN2 exon 7 splicing, and the specificity of splicing-modifying compounds.
01 · Understanding the area
Molecular context for GAYAR's SMA work
Three concepts frame the work here: biomarkers for measuring disease or treatment response, the SMN2 exon 7 splicing mechanism, and compounds designed to modify that splicing process.
Spinal muscular atrophy is primarily associated with loss of functional SMN1 and reduced production of survival motor neuron protein. SMN2 is a closely related gene, but most SMN2 transcripts exclude exon 7 and produce a less stable truncated protein. Increasing exon 7 inclusion can increase full-length SMN production, making SMN2 splicing an important therapeutic target.
SMN2 splicing mechanism
Most SMN2 transcripts exclude exon 7, producing a less stable truncated SMN protein. Increasing exon 7 inclusion can increase production of full-length SMN protein, which is why this splicing event is an important therapeutic target.
Splice-modifying compounds
Risdiplam is an approved SMN2-directed splicing modifier, while branaplam has been investigated as an SMN2 splice-modifying compound. Both can promote exon 7 inclusion, but they differ in sequence specificity and broader splicing effects.
02 · GAYAR research
GAYAR research in this area
Research focusCompleted
Biomarker Discovery
Published work used filtered gene-expression data, Pearson correlation, weighted gene co-expression network analysis, intracluster connectivity, cosine similarity, and differential expression to rank candidate SMA biomarkers. CTDP1 and M6PR emerged across multiple clusters, with external database checks supporting their detectability in blood; clinical validation is still required.
Research focusCurrent
SMN2 Splicing
SVGI studies 285 engineered SMN2 exon 7 5′ splice-site variants under risdiplam and branaplam. The submitted framework quantifies ΔPSI and replicate-normalized responsiveness, builds baseline co-splicing networks, learns graph geometry with a Graph Attention Network, maps that geometry to a two-dimensional lattice with Gromov–Wasserstein optimal transport, and evaluates local three-channel image patches with a CNN.
Research focusIntended
Transcriptome-Wide Therapeutic Effects
Future work is intended to examine whether SMN2-directed splicing modifiers also alter other transcripts or splicing events, with emphasis on compound specificity, off-target effects, and differences between treatment responses. This remains a research direction rather than a completed GAYAR project.
03 · GAYAR in this area
GAYAR work in this area
GAYAR project
SVGI
A Graph-to-Image Representation Framework for Modeling Splice-Site Responses in SMN2 Exon 7 Variants
SVGI models treatment-induced splicing responses across 285 engineered SMN2 exon 7 5′ splice-site variants by combining ΔPSI, replicate-normalized responsiveness, co-splicing network connectivity, GAT-derived graph geometry, Gromov–Wasserstein grid mapping, and image-based classification. In sealed five-fold out-of-fold evaluation, ROC-AUC was 0.641 for branaplam and 0.717 for risdiplam, with average precision of 0.328 and 0.467.
View projectSelected outputs
Conference PaperPublished
Identifying Biomarkers for Spinal Muscular Atrophy: A System Biology-Based Approach
Published conference paper using gene-expression and systems-biology analyses to rank candidate biomarkers for spinal muscular atrophy. CTDP1 and M6PR emerged as consistent candidates across multiple clusters, while clinical validation remains necessary.
View outputConference PaperSubmitted
A Graph-to-Image Representation Framework for Modeling Splice-Site Responses in SMN2 Exon 7 Variants
Submitted SVGI conference paper integrating ΔPSI, replicate-normalized responsiveness, weighted co-splicing structure, GAT-derived geometry, Gromov–Wasserstein graph-to-grid mapping, and an AlexNet-style CNN to model risdiplam and branaplam responses across 285 SMN2 exon 7 splice-site variants.
View output