Prediction Tool

Genomic Prediction Tool

Single variant or batch VCF file prediction

Research use only. MERGE scores and their labels are research outputs, not clinical classifications or individual disease probabilities. AlphaGenome requires your own API key; MERGE can run without it.

Prediction Results

Variant:

⚠️ Multiple Transcripts Detected:

🧬 MERGE Pathogenicity Prediction

MERGE is the ensemble positive-class score from the coding, splicing, non-coding, or non-SNV model. The positive class is pathogenic, or splice-altering for the splicing panel. Its panels contain no AlphaGenome feature because AlphaGenome's official terms do not permit its outputs to be used as training features for machine-learning models.
MERGE + AlphaGenome average combines MERGE with a label-free AlphaGenome percentile (0.5 × MERGE + 0.5 × AlphaGenome percentile) and uses its own interpretation bands. The percentile is obtained from benchmark scores without labels or a fitted classifier; it is not a calibrated pathogenicity probability. It is shown for splicing and non-coding variants only. The components are the AlphaGenome splicing score and noncoding raw maximum, respectively; their mean chromosome-held-out AUROCs are 0.9563 and 0.9774. For coding and non-SNV variants the average is not offered. Benchmark results and score definitions. If AlphaGenome is unavailable for a splicing or non-coding variant, the average is reported as unavailable.
AlphaGenome can be requested for every variant class when a key is supplied and the API returns a result; it is excluded from MERGE training and feeds the separate average only.
Score labels (Pathogenic, Likely Pathogenic, Uncertain Significance, Likely Benign and Benign) name research score bands. They are not ACMG/AMP classifications. Percentages show scores on a 0–100 scale, not an individual's probability of disease.

ⓘ Deployment limits and reproducibility: evo2_1b_base is not among the computed features on this deployment, so it is median-imputed from the training distribution. carbon uses the same SUM log-likelihood-delta scoring as training, but over an 8 193 bp SNV context window instead of the training's 24 576 bp (the larger window exceeds the 24 GB GPU memory). Evo 2 SNV windows also differ from the benchmark by one base. These deployment differences, missing features, service failures and software versions can affect agreement with benchmark scores. Missing ensemble inputs are shown as imputed; missing AlphaGenome values leave the average unavailable.

AlphaGenome Results

HyenaDNA Results

NT Model Results

AlphaMissense Results

ESM-1b Results

GPN-MSA Results

popEVE Results

Evo2 (Local) Results

Evo2 queries the local model with the upstream genomic context to obtain scores. LLR = Reference Score - Variant Score.

Enformer Results

GENERATOR Results

GENERATOR-v2 Results

NT-v2 Results

Carbon Results

Carbon-3B, served by the local GPU gateway (SNV + non-SNV).

✓ Copied to clipboard!

Batch Prediction Results

Total Variants Processed

Total Transcripts Rendered

Variant (Chr:Pos Ref>Alt)	Transcript	Gene	🧬 MERGE Score	MERGE + AG average	AlphaGenome	HyenaDNA	NT Score	AlphaMissense	ESM-1b	GPN-MSA	popEVE	Evo2	Enformer	GENERATOR	GENERATOR-v2	NT-v2

Variant	LiftOver Note	Raw Min	Raw Mean	Raw Max	Quantile Min	Quantile Mean	Quantile Max	Splicing Score

Variant	LLR Score	Interpretation

Variant	Score (L2)	Interpretation

Variant	Transcript ID	Gene	Score	Prediction

Variant	Transcript ID	Gene	Score	Prediction

Variant	LiftOver Note	Score

Variant	LiftOver Note	Score

Variant	LLR Score	Ref Score	Var Score	Interpretation	Context (bp)

Variant	Score	Interpretation	Score Type

Variant	Score	Interpretation	Score Type

Variant	Score (L2)	Interpretation	Score Type

Variant	LiftOver Note	AbsMean	Mean	Std	Max Abs	Top5 AbsMean	Pos Sum	Neg Sum