Toward Advanced Quantitative Prediction of Missense Variant Effects.

Two-stage interpretable model: mutation tolerance of the position, then deviation introduced by the specific substitution

QAFIsplit is a redesign of the Quantitative Assessment of Functional Impact (QAFI) framework that models missense variant effect in two interpretable stages: the mutation tolerance of a position, and the deviation introduced by a specific substitution. Framed as the prediction of a molecular endophenotype, it reaches state-of-the-art quantitative accuracy on independent deep mutational scanning (DMS) benchmarks, surpassing AlphaMissense in most protein domains. In clinical variant classification, it performs on par with established clinical tools without ever being trained on clinical labels, and complements meta-predictors without penalty. QAFI-based strategies were submitted blindly to four independent disease-gene challenges of the CAGI 7 community experiment, backed by strong pre-submission clinical validation (AUC up to 0.99).

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

The main appeal is interpretability: separating what depends on the position from what depends on the substitution helps justify an in silico criterion rather than relying on an opaque score. Not training on clinical labels limits the circularity risk of predictors calibrated on ClinVar. Still, performing on par with established tools in clinical classification is not a gain, and the blinded CAGI 7 results are not reported in the abstract.

Analysis by Dr Thibaut Benquey