Fig. 2: Variant classification using ProteoCast.
From: Proteome-wide prediction of the functional impact of missense variants with ProteoCast

A. Full single-mutational landscape computed for Yorkie protein (yki gene, FlyBase proteoform id: FBpp0288697), a key downstream effector in the Hippo/SWH (Sav/Wts/Hpo) signalling pathway. Each cell of the matrix contains a numerical estimate of the functional impact of an amino acid substitution (y-axis) at a given residue position (x-axis) on the protein function. Darker colours indicate stronger effects, and white cells correspond to the wild-type amino acids. The inverted T-shaped symbol at the N-terminus marks low-confidence predictions (Residue Confidence). The highlighted mutations provoke developmental lethality—as referenced in FlyBase (P65L and Y259N, in red) or occur in DEST2 (N169T, and Q177P, in blue). The pLDDT (predicted local distance difference test) score, shown as a colour bar below the matrix, provides a confidence measure of the AlphaFold2-predicted 3D structure, orange: very low, yellow: low, light blue: high, dark blue: very high. Residues 241–274 and 310–343 correspond to the well-characterised and well-structured WW domains. B ProteoCast classification of Yorkie mutations, where the pale pink and deep purple vertical lines indicate the neutral/mild and mild/impactful frontiers, respectively. All 18 known mutations in our benchmark are indicated with red dots for lethal, blue triangles for DEST2, and blue squares for DGRP. C Distributions of Z-scores for DGRP and DEST2 polymorphisms, as well as hypomorph and lethal mutations annotated in FlyBase, depending on whether they are classified as impactful, mild or neutral. We computed the Z-scores with respect to the distribution of estimated effects for the neutral class within each protein. Three mutations with scores below -25 were excluded from the plot for graphical representation clarity. Source data are provided as a Source Data file.