SpliceAI reports four predicted splice changes for a variant: acceptor gain (DS_AG), acceptor loss (DS_AL), donor gain (DS_DG), and donor loss (DS_DL). The headline delta score is the largest of those four values, from 0 to 1. Its matching delta-position field (DP_AG, DP_AL, DP_DG, or DP_DL) shows where that predicted change lies relative to the variant.
A score is a prediction of splice-altering potential, not a diagnosis and not a direct probability that a variant causes disease. Correct interpretation requires the score type, its paired position, the reference assembly, transcript annotation, raw or masked setting, distance window, and the other clinical evidence for the variant.
As an Amazon Associate I earn from qualifying purchases.
What SpliceAI delta scores measure
SpliceAI is a machine-learning model that predicts whether a DNA variant may alter pre-mRNA splicing. Its output separates two kinds of splice sites—acceptors and donors—and two directions of change—gain and loss.
| Field | Meaning | Question it addresses |
|---|---|---|
| DS_AG | Delta score for acceptor gain | Could the variant create or strengthen an acceptor site? |
| DS_AL | Delta score for acceptor loss | Could the variant weaken or remove an acceptor site? |
| DS_DG | Delta score for donor gain | Could the variant create or strengthen a donor site? |
| DS_DL | Delta score for donor loss | Could the variant weaken or remove a donor site? |
The project defines the variant delta score as max(DS_AG, DS_AL, DS_DG, DS_DL). The result ranges from 0 to 1. The Illumina SpliceAI repository says this value “can be interpreted as the probability of the variant being splice-altering.” That wording describes the model’s prediction of splice impact; it does not mean a 0.5 score is a 50% chance of pathogenicity, cancer, or any other clinical outcome.
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
Always retain the suffix. A donor-loss prediction is biologically different from an acceptor-gain prediction, even when their numeric scores are identical.
How to read delta positions (DP)
Each score has a position field with the same suffix: DP_AG, DP_AL, DP_DG, or DP_DL. The position belongs to that specific prediction. If DS_DG is the largest score, read DP_DG—not DP_AG or a generic position field—to locate the predicted donor change.
In the convention documented by the SpliceAI repository, positive positions are downstream of the variant and negative positions are upstream. A DP value is an offset relative to the variant representation, not an absolute chromosome coordinate.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Worked example: DS_DG 0.91 and DP_DG −2
The repository’s documented RYR1 output contains DS_DG 0.91 and DP_DG −2. Under its sign convention, the model predicts a strong donor-gain signal at a site two bases upstream of the variant. This tells you the predicted event type and relative location. It does not, by itself, prove which exon or intron will appear in the mature transcript, nor does it establish that the variant is pathogenic.
Different databases can display SpliceAI-derived fields with their own formatting or conventions. Confirm the software and output specification before transferring a sign interpretation from a database manual to raw SpliceAI output.
What a SpliceAI score of 0.5 means
In the original project documentation, the approximate operating points are:
| Delta-score range or cutoff | How the project characterizes it | Appropriate use |
|---|---|---|
| 0.2 | High recall | Prioritize variants so fewer possible splice-altering variants are missed. |
| 0.5 | Recommended cutoff | General prioritization in the project’s evaluation framework. |
| 0.8 | High precision | Prioritize a smaller set with stronger predicted splice impact. |
Those labels describe model operating trade-offs, not benign, uncertain, or pathogenic categories. A score of 0.5 means the model assigns substantial splice-altering signal according to its calibration; it does not mean “50% pathogenic.” A score below 0.2 can still matter in a gene where a small splice change has a known mechanism, and a score above 0.8 still needs clinical corroboration.
Why thresholds differ between analyses
Thresholds used for clinical evidence are tied to the data, annotation, transcript set, and distance window used to derive them. ClinGen’s Splicing Subgroup response to feedback (updated March 22, 2024) describes more than one calibration context rather than one universal SpliceAI scale.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
ClinGen example using raw scores
One analysis used raw SpliceAI delta scores, a maximum distance of ±4,999 nucleotides, and cutoffs of at most 0.1 and at least 0.2. In that framework, the proposed generic computational evidence use is PP3 at or above 0.2 and BP4 at or below 0.1. These are calibrated evidence recommendations, not diagnoses. Scores between the cutoffs require other evidence and should not be forced into either category.
ClinGen example using precomputed hg19 scores
A separate analysis used precomputed hg19 scores, a ±50-nucleotide distance, and cutoffs below 0.1 and at least 0.5. The same document notes that precomputed scores based on GENCODE v24 can differ from lookup scores based on GENCODE v43. Therefore, a threshold copied from one analysis cannot safely be applied to an output generated with different annotations or settings.
ClinGen’s calibration descriptions include datasets of 1,008 BRCA1/BRCA2 variants, 659 mismatch-repair-gene variants, 284 NF1 variants, and 1,070 POU1F1 variants. Those numbers describe analyzed calibration sets, not disease prevalence and not universal performance for every gene or variant class.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallA step-by-step way to read a SpliceAI annotation
- Identify the exact variant representation. Check chromosome, position, reference and alternate alleles, genome assembly, and normalization. An assembly or allele mismatch can change the result or prevent scoring.
- Read all four DS fields. Record DS_AG, DS_AL, DS_DG, and DS_DL rather than looking only at a headline value. Comparing the four values tells you which event type dominates.
- Find the maximum. Calculate or verify
max(DS_AG, DS_AL, DS_DG, DS_DL). This is the headline delta score. - Keep the matching DP field. If DS_AL is largest, pair it with DP_AL. State whether the position is upstream or downstream using the convention documented for that output.
- Inspect the transcript and gene context. Determine which transcript was annotated, whether the predicted site lies near an exon boundary, and whether the gene’s disease mechanism is known to involve abnormal splicing.
- Record configuration and provenance. Save the software or database version, reference assembly, annotation release, raw or masked mode, and maximum distance. Two apparently different scores may simply come from different configurations.
- Apply a calibrated threshold only in its matching context. Do not mix the original project’s 0.2/0.5/0.8 operating points with ClinGen PP3 or BP4 recommendations without stating which framework and settings you are using.
- Combine with independent evidence. Use RNA data, segregation, population frequency, functional assays, phenotype, established gene-disease mechanism, and other applicable evidence before making a clinical classification.
Inputs and settings that affect the result
The documented command-line workflow takes a VCF, a reference-genome FASTA file, and a gene annotation. Built-in annotation options use GENCODE V24 canonical files for GRCh37 or GRCh38. The documented default maximum distance is 50 bases. The program is described as scoring SNVs and simple indels of the supported forms within annotated genes.
| Setting or input | Why it matters |
|---|---|
| Reference assembly | GRCh37 and GRCh38 have different coordinates and sequence contexts; scores are not interchangeable by position alone. |
| Transcript and gene annotation | Canonical transcript choices and annotation releases determine which splice sites are evaluated. |
| Maximum distance | The window controls how far from the variant a predicted site may be reported. The default documented value is 50, while ClinGen examples use ±50 or ±4,999 nucleotides in different analyses. |
| Variant normalization | Representation of indels and alleles affects the sequence presented to the model. |
| Raw versus masked mode | The two modes intentionally answer different interpretation questions and can produce different values. |
Raw and masked SpliceAI files are not interchangeable
The repository distinguishes raw (-M 0) and masked (-M 1) predictions.
- Raw scores: include some predictions that strengthen annotated splice sites and weaken unannotated sites. The project recommends these for alternative-splicing analysis.
- Masked scores: set those predictions to zero. The project recommends masked scores for variant interpretation.
When comparing results, write down the mode explicitly. A raw score of zero and a masked score of zero do not necessarily carry the same information, and a score obtained in one mode should not be substituted for a clinical threshold calibrated in the other.
When SpliceAI returns no score
A missing annotation is not the same as a predicted score of zero. The documented software lists several reasons a variant may not be scored:
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minute- The variant is outside the genes in the selected annotation.
- It lies within 5 kilobases of a chromosome end.
- A deletion is longer than twice the configured maximum distance.
- The reference allele does not match the supplied FASTA.
- The variant is not one of the supported SNV or simple-indel forms.
Check the VCF allele, chromosome naming, assembly, FASTA index, annotation release, and distance setting before interpreting an absent field. If the variant is genuinely outside the selected annotation, the result says that this run could not evaluate it; it does not establish no splice effect.
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
Does a high SpliceAI score mean a variant is pathogenic?
No. SpliceAI predicts a molecular consequence—potential alteration of splice-site usage. Clinical classification asks a broader question: whether the variant causes disease in a particular gene and clinical context. That decision can depend on the gene-disease mechanism, transcript relevance, population data, phenotype, segregation, functional assays, RNA studies, and other validated evidence.
A high score can justify follow-up, such as checking for an exon-skipping assay or reviewing RNA evidence when an appropriate specimen is available. It cannot establish the exact transcript consequence from the four summary values alone. For example, the scores do not by themselves prove exon skipping, intron retention, or creation of a pseudoexon.
Common interpretation mistakes and fixes
Calling DS_AG “the SpliceAI score”
DS_AG is only one directional component. Compare all four DS fields and report the maximum together with its event type.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Reading DP as a genomic coordinate
DP is a relative offset. Use the variant’s normalized representation and the documented sign convention to map it to sequence context.
Ignoring the suffix when pairing fields
Pair DS_DG with DP_DG, DS_AL with DP_AL, and so on. A position from another suffix describes a different prediction.
Applying 0.5 as a disease probability
The score concerns predicted splice alteration, not pathogenicity. Use a threshold only within the framework and configuration that support it.
Treating an absent field as zero
Investigate annotation coverage, chromosome-end distance, deletion size, allele/FASTA consistency, and supported variant type.
Comparing scores generated with different annotations
Record GENCODE release, assembly, transcript set, distance window, and raw or masked mode before deciding that two results conflict.
Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
Comparing two SpliceAI annotations
When two databases or pipelines report different values, first assume the inputs or settings may differ rather than treating one value as an error. Compare the reference assembly and normalized alleles, then identify the transcript and gene annotation release. A GRCh37 result cannot be compared by coordinate alone with a GRCh38 result, and GENCODE V24 canonical annotations may produce different values from a lookup based on GENCODE V43.
Next compare the maximum distance window. A run limited to 50 nucleotides can miss or omit a site that is eligible in a run using a ±4,999-nucleotide window. Finally, check whether each result is raw or masked and whether you are comparing the same directional field. A headline maximum can be identical while the winning event type or DP offset differs, so retain all four DS and DP pairs when auditing a discrepancy.
| Comparison axis | Details to record |
|---|---|
| Assembly and normalization | GRCh37 or GRCh38, chromosome naming, allele representation, and normalized indel form. |
| Annotation | Gene/transcript set, canonical-transcript choice, and release such as GENCODE V24 or V43. |
| Prediction mode | Raw -M 0 or masked -M 1. |
| Distance | Configured maximum, including whether the analysis uses 50 or ±4,999 nucleotides. |
| Fields | All four DS values and each matching DP value, not only the displayed maximum. |
What the four values can and cannot tell you
The four directional values summarize predicted changes in donor and acceptor usage. They help rank variants for review and identify the type and relative location of the strongest signal. They do not provide a complete transcript model. Even a high donor-gain value does not specify, from the summary output alone, whether the resulting transcript will contain a skipped exon, retained intron, or pseudoexon.
Recommended Free Tools
If two directional scores are close, report both rather than hiding the second value. The maximum still determines the headline score, but the complete vector preserves information for later transcript review and makes the interpretation reproducible. A low maximum is not proof of normal splicing, especially when the relevant transcript is absent from the chosen annotation or the variant falls outside the configured window.
Practical reporting template
A reproducible report should state the variant, assembly, annotation, score mode, distance window, four DS values, the maximum, and the matching DP value. For example:
Variant: [normalized variant, assembly stated]
Annotation: [gene/transcript and annotation release]
Mode: raw (-M 0) or masked (-M 1)
Maximum distance: [value]
DS_AG: [value] DP_AG: [value]
DS_AL: [value] DP_AL: [value]
DS_DG: [value] DP_DG: [value]
DS_DL: [value] DP_DL: [value]
Headline: [largest DS], predicting [gain/loss and donor/acceptor] at [matching DP offset]
Interpretation: predicted splice impact; not a standalone clinical classificationThis format prevents a headline score from being detached from the event type, position, and computational provenance needed to reproduce it. Include the command-line or database source when available, and preserve the original unrounded values if a threshold decision depends on a boundary such as 0.1 or 0.2.
Software and licensing note
The SpliceAI repository states that its source code is under a PolyForm Strict License and that trained models are provided for academic and noncommercial use. Commercial use requires a commercial license; the README identifies Illumina as the licensing contact. Confirm the current license terms before embedding the model in a commercial product or service.
Bottom line
Interpret SpliceAI in pairs: first identify the largest of DS_AG, DS_AL, DS_DG, and DS_DL, then read the DP field with the same suffix and apply the documented upstream/downstream convention. Treat the value as a calibrated prediction of splice-altering potential. A score of 0.5 is not a 50% disease risk, a score of 0.9 is not a diagnosis, and no score—or a missing score—can replace transcript-aware, configuration-aware clinical evidence.
Frequently Asked Questions
What does a SpliceAI score of 0.5 mean?
It is a substantial predicted splice-altering signal and the original project describes 0.5 as a recommended operating cutoff. It is not a 50% probability that the variant is pathogenic. Clinical evidence frameworks may use different cutoffs tied to specific annotations, distance windows, and score modes.
How do I read DS_AG and DP_AG together?
DS_AG is the predicted strength of acceptor gain. DP_AG is the relative offset for that same prediction. Report them as a pair, and use the sign convention documented for the output: in the SpliceAI repository, positive values are downstream and negative values are upstream.
What does a negative delta position mean?
For raw SpliceAI output following the repository convention, a negative DP means the predicted splice-site change is upstream of the variant. DP is an offset, not an absolute genomic coordinate, and database displays may use different conventions.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Does a high SpliceAI score prove pathogenicity?
No. It predicts possible alteration of splice-site usage. Pathogenicity assessment also requires the relevant transcript and gene-disease mechanism plus evidence such as population data, phenotype, segregation, RNA studies, or functional assays.
Why might SpliceAI show no score?
The variant may be outside the selected annotation, within 5 kb of a chromosome end, a deletion may exceed twice the configured distance, the reference allele may not match the FASTA, or the variant may not be a supported SNV or simple indel. A missing field should not be converted to a score of zero.
Are raw and masked SpliceAI scores interchangeable?
No. Raw mode includes some predictions that strengthen annotated sites or weaken unannotated sites; masked mode sets those predictions to zero. The repository recommends raw scores for alternative-splicing analysis and masked scores for variant interpretation, so record the mode before applying any threshold.
Why do two tools report different SpliceAI values for one variant?
Differences can result from assembly or normalization, transcript and GENCODE release, raw versus masked mode, maximum distance, or the specific DS/DP fields being compared. Document all of these settings before concluding that the annotations conflict.
Free tools Windows power users keep installed
One-click scans. No signup required.
The Bottom Line
Use the largest DS value to identify the predicted event, pair it with the DP field carrying the same suffix, and document assembly, annotation, distance, and raw or masked mode. SpliceAI prioritizes possible splice effects; it does not independently establish a disease diagnosis.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
