Calibration of AlphaGenome Variant Impact and SpliceAI2 scores as PP3/BP4 evidence for missense, non-coding, and synonymous variants

Calibration of AlphaGenome Variant Impact and SpliceAI2 scores into PP3/BP4 evidence levels, by variant class

The authors calibrate the AlphaGenome Variant Impact score (AVI) for missense variants and for six non-coding and synonymous variant classes, and SpliceAI2 for these six classes, for use as graded PP3/BP4 evidence. For missense variants, AVI reached evidence between moderate and strong in both directions but did not outperform its component AlphaMissense. AVI reached strong pathogenic evidence in the three classes adjacent to exons, moderate evidence in proximal introns and at other synonymous sites, and only supporting evidence, at an unstable threshold, in deep introns; SpliceAI2 reached the same levels, except one level lower at near-canonical positions and none in deep introns. No reached level failed validation on later ClinVar classifications, although 14 of 30 benign levels were not tested. Thresholds depend on the choice of controls, ClinVar benign classifications or population variants: ClinVar likely benign classifications changed about 60-fold in number at 20 nucleotides from the exon without a change in measured splicing.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

Preprint, not peer reviewed. This is the missing piece for using these new scores in a defensible ACMG classification: thresholds per variant class, checked against later classifications. The most useful observation is that thresholds depend on the choice of controls: the ClinVar step at 20 nucleotides from the exon reflects a classification convention rather than biology, which calls for caution with thresholds calibrated on ClinVar benign variants in introns.

Analysis by Dr Thibaut Benquey