Fig. 1

Algorithm. In our algorithm mp_lipid_acc, the protein is cut into horizontal slices (a). In each slice, the coordinates of the Cα atoms are projected onto the xy-plane, from which points first the convex hull, then the concave hull and then the ‘concave shell’ are computed ((b) and see implementation). For β-barrels, the center-of-mass in each slice is computed from the coordinates of the Cα atoms and only residues which are part of the concave shell and have a Cα-Cβ-COM angle larger than a cutoff value of 65° are classified as lipid accessible (c). For instance, the residue with a sidechain orientation represented by the yellow vector in (c) would be lipid accessible, whereas the sidechain in blue would be lipid inaccessible