Compressed representation of natural image variability underlies high accuracy in face recognition

Accurately recognising faces is fundamental to human social interaction. In recent years it has become clear that people’s accuracy differs markedly depending on viewer’s familiarity with a face and their individual skill, but the cognitive and neural bases of these accuracy differences are not understood. We examined cognitive representations underlying these accuracy differences by measuring similarity ratings to natural facial image variation. Using image averaging, and inspired by the computation of Analysis of Variance, we partitioned image variation into differences between faces (between-identity variation) and differences between photos of the same face (within-identity variation). Contrary to prevailing accounts of human face recognition and perceptual learning, we found that modulation of within-identity variation – rather than between-identity variation – was associated with high accuracy. First, similarity of within-identity variation was compressed for familiar faces relative to unfamiliar faces. Second, viewers that are extremely accurate in face recognition – ‘super-recognisers’ – showed enhanced compression of within-identity variation that was most marked for familiar faces. We also present computational analysis showing that cognitive transformations of between- and within-identity variation make separable contributions to perceptual expertise in unfamiliar and familiar face identification respectively. We conclude that inter- and intra-individual accuracy differences primarily arise from differences in the representation of familiar face image variation.

PCD Detect: enhancing ciliary features through image averaging and classification

Primary ciliary dyskinesia (PCD) is an inherited disorder of the motile cilia. Early accurate diagnosis is important to help prevent lung damage in childhood and to preserve lung function. Confirmation of a diagnosis traditionally relied on assessment of ciliary ultrastructure by transmission electron microscopy (TEM); however, >50 known PCD genes have made the identification of biallelic mutations a viable alternative to confirm diagnosis. TEM and genotyping lack sensitivity, and research to improve accuracy of both is required. TEM can be challenging when a subtle or partial ciliary defect is present or affected cilia structures are difficult to identify due to poor contrast. Here, we demonstrate software to enhance TEM ciliary images and reduce background by averaging ciliary features. This includes an option to classify features into groups based on their appearance, to generate multiple averages when a nonhomogeneous abnormality is present. We validated this software on images taken from subjects with well-characterized PCD caused by variants in the outer dynein arm (ODA) heavy chain gene DNAH5. Examining more difficult to diagnose cases, we detected 1) regionally restricted absence of the ODAs away from the ciliary base, in a subject carrying mutations in DNAH9; 2) loss of the typically poorly contrasted inner dynein arms; and 3) sporadic absence of part of the central pair complex in subjects carrying mutations in HYDIN, including one case with an unverified genetic diagnosis. We show that this easy-to-use software can assist in detailing relationships between genotype and ultrastructural phenotype, improving diagnosis of PCD.

Binding site localization on non-homogeneous cell surfaces using topological image averaging

Antibody binding to cell surface proteins plays a crucial role in immunity and the location of an epitope can altogether determine the immunological outcome of a host-target interaction. Techniques available today for epitope identification are costly, time-consuming, and unsuited for high-throughput analysis. Fast and efficient screening of epitope location can be useful for the development of therapeutic monoclonal antibodies and vaccines. In the present work, we have developed a method for imaging-based localization of binding sites on cellular surface proteins. The cellular morphology typically varies, and antibodies often bind in a non-homogenous manner, making traditional particle-averaging strategies challenging for accurate native antibody localization. Nanometer-scale resolution is achieved through localization in one dimension, namely the distance from a bound ligand to a reference surface, by using topological image averaging. Our results show that this method is well suited for antibody binding site measurements on native cell surface morphology.

Protein–lipid architecture of a cholinergic postsynaptic membrane

The cholinergic postsynaptic membrane is an acetylcholine receptor-rich membrane mediating fast chemical communication at the nerve–muscle synapse. Here, cryo-EM is used to examine the protein–lipid architecture of this membrane in tubular vesicles obtained from the (muscle-derived) electric organ of the Torpedo ray. As reported earlier, the helical arrangement of the protein component of the vesicles facilitates image averaging and enables us to determine how cholesterol and phospholipid molecules are distributed in the surrounding matrix, using headgroup size as a means to discriminate between the two kinds of lipid. It is shown that cholesterol segregates preferentially around the receptors in both leaflets of the lipid bilayer, interacting robustly with specific transmembrane sites and creating a network of bridging microdomains. Cholesterol interactions with the receptor are apparently essential for stabilizing and maintaining its physiological architecture, since the transmembrane structure contracts, involving displacements of the helices at the outer membrane surface by ∼2 Å (1–3 Å), when this lipid is extracted. The microdomains may promote cooperativity between neighbouring receptors, leading to an enhanced postsynaptic response.