International audienceVisual understanding is often based on measuring similarity between observations. Learning similarities specific to a certain perception task from a set of examples has been shown advantageous in various computer vision and pattern recognition problems. In many important applications, the data that one needs to compare come from different representations or modalities, and the similarity between such data operates on objects that may have different and often incommensurable structure and dimensionality. In this paper, we propose a framework for supervised similarity learning based on embedding the input data from two arbitrary spaces into the Hamming space. The mapping is expressed as a binary classification problem with positive and negative examples, and can be efficiently learned using boosting algorithms. The utility and efficiency of such a generic approach is demonstrated on several challenging applications including cross-representation shape retrieval and alignment of multi-modal medical images

PtdIns(3,5)P(2) is required for cargo-selective sorting to the vacuolar lumen via the multivesicular body (MVB). Here we show that Ent3p, a yeast epsin N-terminal homology (ENTH) domain-containing protein, is a specific PtdIns(3,5)P(2) effector localized to endosomes. The ENTH domain of Ent3p is essential for its PtdIns(3,5)P(2) binding activity and for its membrane interaction in vitro and in vivo. Ent3p is required for protein sorting into the MVB but not for the internalization step of endocytosis. Ent3p is associated with clathrin and is necessary for normal actin cytoskeleton organization. Our results show that Ent3p is required for protein sorting into intralumenal vesicles of the MVB through PtdIns(3,5)P(2) binding via its ENTH domain.

Integration of the human immunodeficiency virus (HIV-1) cDNA into the human genome is catalysed by integrase. Several studies have shown the importance of the interaction of cellular cofactors with integrase for viral integration and infectivity. In this study, we produced a stable and functional complex between the wild-type full-length integrase (IN) and the cellular cofactor LEDGF/p75 that shows enhanced in vitro integration activity compared with the integrase alone. Mass spectrometry analysis and the fitting of known atomic structures in cryo negatively stain electron microscopy (EM) maps revealed that the functional unit comprises two asymmetric integrase dimers and two LEDGF/p75 molecules. In the presence of DNA, EM revealed the DNA-binding sites and indicated that, in each asymmetric dimer, one integrase molecule performs the catalytic reaction, whereas the other one positions the viral DNA in the active site of the opposite dimer. The positions of the target and viral DNAs for the 3 0 processing and integration reaction shed light on the integration mechanism, a process with wide implications for the understanding of viral-induced pathologies.