Discovery & Browsing
One score to rank the whole proteome — and three lenses to explore it.
With well over a hundred thousand isoforms in the knowledgebase, the hard part isn't computing consequences — it's finding the ones that matter. The Discovery Index pre-computes a single Structural Impact Score (SIS) for every isoform, so the entire proteome can be ranked and filtered in real time to surface the highest-impact splice events.
The Structural Impact Score (SIS)
The SIS is a normalised 0–100 composite of five biological axes. Each axis is mapped to a sub-score between 0 and 1 with a saturating function, then combined with fixed weights that sum to 100. Because every axis is bounded, no single event type can dominate the ranking, and the number is directly interpretable — an isoform scoring 62 carries roughly 62 % of the maximum evidence for structural consequence tracked in the knowledgebase.
Each si ∈ [0, 1] (see the axes below); the weights sum to 100, so base is bounded 0–100.
The model's mean pLDDT scales the score by a factor of 0.70–1.00: low-confidence models are discounted but never zeroed, so a real domain loss on a lower-confidence novel isoform stays visible while genuine noise is down-weighted.
| Axis | Weight | Sub-score (0–1) | What it captures |
|---|---|---|---|
| Domain disruption | 30 | 1 − 0.4^(lost + 0.5·gained) | Loss of a conserved Pfam domain is the strongest signal of functional disruption. The term saturates (1 lost → 0.60, 2 → 0.84), so an isoform that sheds many domains can't dominate the ranking. |
| Length change | 20 | min(|Δlen| / 150, 1) | A proxy for gross structural change, and the only structural signal available for novel isoforms with no annotated Pfam domain. Saturates at a 150-aa change relative to the canonical. |
| Neoantigen priority | 20 | P1→1 · P2→0.75 · P3→0.5 · P4→0.25 | The best neoepitope priority tier per isoform (presentation consensus × immunogenicity) — a P1 hit contributes full weight, a P4 a quarter. |
| NMD fate | 15 | target→1, else→0 | Transcripts predicted to be degraded by nonsense-mediated decay likely never yield a stable protein — a strong regulatory consequence. |
| PTM accessibility | 15 | lost→1 · exposed/buried→0.6 | Deleting a post-translational-modification site outweighs merely repositioning it, where the residue is retained but its solvent accessibility changes. |
Exploring the Knowledgebase
Three complementary lenses open the same data from different angles, each reachable from the site header. They share the underlying models but rank and filter for a different question.
Ranks every isoform proteome-wide by its Structural Impact Score. Filter by impact profile and structural confidence, then open any row for the full isoform page.
Open the explorer →A proteome-wide table of splice-junction neoantigens, ranked into P1–P4 priority tiers with filters for HLA allele, binding, and junction type.
Ranks drugged genes by how much their splicing subtracts from each drug's binding pocket — preserved, disrupted, or lost.
Impact profiles
The Impact Explorer ships with preset filters that answer common questions in one click, on top of the free sliders for structural confidence and isoforms-per-gene: