Documentation / Computational Predictors / SpliceAI
SpliceAI Score Interpretation and ACMG Thresholds
SpliceAI reports donor and acceptor gain or loss. Folklore reads the maximum delta score through explicit thresholds: values at or below 0.1 can support defined benign guards, values from 0.2 enter the PP3_splice path with strength modulation, and PVS1 blocks duplicate computational evidence for the same loss-of-function mechanism.
Review one public GRCh38 variant against Folklore's current evidence snapshot. Public search accepts a variant expression only. Do not enter patient or case information.
Search a public variantWhat the Four Scores Represent
SpliceAI is a deep learning model developed by Illumina that predicts changes in splice-site usage from primary sequence. Splicing is the process by which the cell removes non-coding sections (introns) from the pre-mRNA and joins the coding sections (exons) to produce the final messenger RNA. A predicted change is computational evidence, not confirmation of a transcript or protein consequence.
The model was published in Cell (Jaganathan et al., 2019) and is widely used in clinical genetics laboratories. It is one of two computational tools in Folklore that directly influences ACMG classification.
Four Delta Scores
SpliceAI produces four model scores, each representing a predicted change in splice-site usage. The values range from 0 to 1 but are not calibrated probabilities of a splice event. Folklore uses the maximum of the four delta scores for its computational-evidence thresholds.
| Score | Name | Meaning |
|---|---|---|
| DS_AG | Acceptor Gain | The model predicts increased acceptor usage at a nearby position |
| DS_AL | Acceptor Loss | The model predicts reduced usage of an existing acceptor |
| DS_DG | Donor Gain | The model predicts increased donor usage at a nearby position |
| DS_DL | Donor Loss | The model predicts reduced usage of an existing donor |
Clinical Score Thresholds
| Max Score Range | Interpretation |
|---|---|
| 0.0 to 0.1 | Can support defined benign splice guards when the consequence is eligible. |
| Above 0.1 to below 0.2 | Indeterminate for the formal splice evidence path. |
| 0.2 to below 0.5 | PP3_splice Supporting when all eligibility guards pass. |
| 0.5 to below 0.8 | PP3_splice Moderate when all eligibility guards pass. |
| 0.8 to 1.0 | PP3_splice Strong when all eligibility guards pass. |
How Folklore Uses SpliceAI
SpliceAI scores are used in three distinct ways within the classification engine. Each role has a specific threshold and clinical rationale.
| Condition | ACMG Effect | Rationale |
|---|---|---|
| 0.2 to below 0.5 | PP3_splice Supporting | Supporting splice evidence when consequence, PVS1 and mutual-exclusion guards pass. |
| 0.5 to below 0.8 | PP3_splice Moderate | Moderate splice evidence when the same eligibility guards pass. |
| 0.8 or higher | PP3_splice Strong | Strong splice evidence when disease-association and other eligibility guards pass. |
| Max score below 0.1 or absent | Guard for missense BP4 | BayesDel benign evidence is blocked when a concerning splice signal is present. |
| Max score <= 0.1 | BP7 and BP4_splice guards | Benign splice evidence is consequence-specific. Canonical splice-motif and other ineligible consequences remain excluded. |
PVS1 Double-Counting Guard
When PVS1 (loss-of-function) is triggered for a variant, PP3_splice is not applied. This prevents counting the same predicted splice-related loss-of-function mechanism as both PVS1 and PP3 evidence. Pure missense variants with BayesDel data but no VEP splice consequence do not receive a second computational splice criterion. When both missense and splice consequences are present, the classifier uses mutual-exclusion safeguards so correlated predictions do not become additive votes.
Score Source
Folklore reads SpliceAI scores from a managed precomputed GRCh38 dataset rather than calculating them during each analysis. The dataset records its source files and version so a result can be interpreted against the evidence snapshot used for that run. Asset updates can change later results and must not be conflated with within-snapshot repeatability.
Limitations
SpliceAI is a sequence-based computational model. It does not establish which transcript is expressed in the relevant tissue, whether a predicted transcript change occurs, or what that change means for disease. Coverage is bounded by the managed precomputed file set, genome build, and transcripts represented in that asset. A missing score means unavailable computational evidence, not evidence of no splice effect. Appropriate RNA evidence can test a predicted effect, but assay design, tissue, and interpretation require qualified review.
Interpretation Boundary
A high model score can contribute computational evidence only when the classifier's consequence, disease-association, PVS1, and mutual-exclusion safeguards pass. It does not confirm aberrant splicing, establish pathogenicity, or determine whether an assay is clinically indicated.
For the managed asset, field provenance, genome build, and coverage boundary, see SpliceAI precomputed scores. For how computational evidence combines with other criteria, see the ACMG/AMP framework.
Reference: Jaganathan K, et al. Cell. 2019;176(3):535-548. PMID: 30661751
ClinGen SVI: Walker LC, et al. Am J Hum Genet. 2023;110(7):1046-1067. PMID: 37352859