The GPCRVP score reliably predicts the impact of GLP1R human variants on receptor function.

GLP1R

Structural and dynamic descriptors extracted from molecular dynamics simulations of GLP1R complexes, combined with AlphaMissense and REVEL scores

The GLP-1 receptor, encoded by GLP1R, regulates glucose homeostasis and body weight and is a major therapeutic target in type 2 diabetes and obesity; missense variants in the gene may alter both disease risk and treatment response. The authors built the GPCRVP score by combining molecular dynamics simulations of wild-type and mutant GLP1R complexes with the AlphaMissense and REVEL predictors. Descriptors extracted from the simulations capture membrane context, allosteric communication and mutation-induced contact rearrangements; the model was trained on 58 experimentally characterised GLP1R variants. GPCRVP showed strong agreement with experimental data and an approximately 78% reduction in false-positive predictions compared with AlphaMissense on independent validation sets. Structural models were generated for all 174 known human GLP1R missense variants outside the training and validation sets.

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

Specializing a predictor by receptor family rather than applying it across the proteome is probably the right direction, and cutting false positives is the axis that matters: AlphaMissense's over-calling is what currently limits its use in classification. Training on 58 variants remains very narrow, and nothing is shown outside GLP1R — the generic "GPCR" claim is not demonstrated here. No treatment response data are reported, so no pharmacogenetic use at this stage.

Analysis by Dr Thibaut Benquey