Oct 30, 2026
This research could help in understanding genetic testing results for patients with hereditary hemorrhagic telangiectasia and pulmonary arterial hypertension. It provides a better way to identify potentially harmful variants, which can guide treatment decisions. Patients might discuss these findings with their healthcare providers as new approaches in genetic testing.
The study aims to improve how we classify genetic changes (variants) in the ACVRL1 gene, which is linked to hereditary hemorrhagic telangiectasia (HHT) and pulmonary arterial hypertension. This is important because some variants are not clearly understood, making it hard to know if they are harmful or not.
The new models achieved high accuracy in classifying variants, reaching scores that are almost equivalent to AlphaMissense, which is currently one of the best tools for this task. However, the models had difficulty with benign variants, showing some limitations.
Researchers looked at 899 ACVRL1 variants by assessing their protein stability, conservation (how much a genetic feature is maintained across species), and structure. They created computer models to classify these variants based on 311 well-documented cases, comparing their effectiveness to existing classifiers like AlphaMissense.
The study is based on a computer modeling approach and uses previously classified variants for validation. This means it may not cover all types of variants or situations. It also relies on existing databases and knowledge, and the classification should be considered as hypotheses needing further testing, rather than definitive evidence.
For your doctor:
PMID: 43027989
DOI: 10.3390/ijms27219742
Missense variants in ACVRL1 (ALK1) cause hereditary hemorrhagic telangiectasia type 2 (HHT2), heritable pulmonary arterial hypertension (PAH), or both in the same patient, and many remain variants of uncertain significance. AlphaMissense classifies them accurately but does not show which evidence drives a call. I annotated 899 ACVRL1 missense variants with protein-stability estimates (FoldX, DynaMut2), a GREMLIN-style conservation score, and structural context. I then trained random-forest classifiers on 311 variants with labels traceable to ClinVar or a publication (296 pathogenic-like, 15 benign-like). A structural/conservation-only classifier (Model 1b) reached a cross-validated ROC-AUC of 0.954 (95% CI 0.916-0.984), and a model adding sequence-based predictors (Model 1) reached 0.979, versus 0.983 for AlphaMissense on the same variants. On variants first reported in 2025-2026, Model 1b scored 0.976, Model 1 0.992, and AlphaMissense 0.992; equivalence within ±0.05 AUC was supported for Model 1 only. Benign variants were the main weakness: Model 1b correctly classified 10/15, against 14/15 for AlphaMissense. Conservation drove most calls, with predicted destabilization contributing to a minority. An interpretable, gene-specific classifier can approach AlphaMissense for ACVRL1 while showing the structural or evolutionary evidence behind each call. Its calls are hypotheses, not functional evidence, and can prioritize VUS for functional testing.