Numerous algorithms have been proposed for transferring knowledge from a label-rich domain (source) to a label-scarce domain (target). Almost all of them are proposed for closed-set scenario, where the source and the target domain completely share the class of their samples. We call the shared class the "known class." However, in practice, when samples in target domain are not labeled, we cannot know whether the domains share the class. A target domain can contain samples of classes that are not shared by the source domain. We call such classes the "unknown class" and algorithms that work well in the open set situation are very practical. However, most existing distribution matching methods for domain adaptation do not work well in this setting because unknown target samples should not be aligned with the source. In this paper, we propose a method for an open set domain adaptation scenario which utilizes adversarial training. A classifier is trained to make a boundary between the source and the target samples whereas a generator is trained to make target samples far from the boundary. Thus, we assign two options to the feature generator: aligning them with source known samples or rejecting them as unknown target samples. This approach allows to extract features that separate unknown target samples from known target samples. Our method was extensively evaluated on domain adaptation setting and outperformed other methods with a large margin in most setting.

Homologous recombination (HR) repairs DNA double-strand breaks (DSBs) to maintain genomic integrity. Recombinase recruited to the DSBs by the mediator protein BRCA2 catalyzes the homology-directed repair. During meiotic HR, programmed DSBs are introduced genome-wide but their repair mechanisms, including the regulation of BRCA2, have remained largely elusive. Here we identify a meiotic localizer of BRCA2, MEILB2/HSF2BP, that localizes to the site of meiotic DSBs in mice. Disruption of Meilb2 abolishes the localization of RAD51 and DMC1 recombinases in spermatocytes, leading to errors in DSB repair and male sterility. MEILB2 directly binds to BRCA2 and regulates its association to meiotic DSBs. We map the MEILB2-binding domain within BRCA2 that is distinct from the canonical DNA-binding domain but is sufficient to localize to meiotic DSBs in a MEILB2-dependent manner. We conclude that localization of BRCA2 to meiotic DSBs is mediated by MEILB2, which is an integral mechanism to repair abundant meiotic DSBs.

During centriole duplication, a single daughter centriole is formed next to the mother centriole. The molecular mechanism that determines a single duplication site remains a long-standing question. Here, we show that intrinsic self-organization of Plk4 is implicated in symmetry breaking in the process of centriole duplication. We demonstrate that Plk4 has an ability to phase-separate into condensates via an intrinsically disordered linker and that the condensation properties of Plk4 are regulated by autophosphorylation. Consistently, the dissociation dynamics of centriolar Plk4 are controlled by autophosphorylation. We further found that autophosphorylated Plk4 is already distributed as a single focus around the mother centriole before the initiation of procentriole formation, and is subsequently targeted for STIL-HsSAS6 loading. Perturbation of Plk4 self-organization affects the asymmetry of centriolar Plk4 distribution and proper centriole duplication. Overall, we propose that the spatial pattern formation of Plk4 is a determinant of a single duplication site per mother centriole.

In each cell cycle, centrioles are duplicated to produce a single copy of each preexisting centriole. At the onset of centriole duplication, the master regulator Polo-like kinase 4 (Plk4) undergoes a dynamic change in its spatial pattern around the preexisting centriole, forming a single duplication site. However, the significance and mechanisms of this pattern transition remain unknown. Using super-resolution imaging, we found that centriolar Plk4 exhibits periodic discrete patterns resembling pearl necklaces, frequently with single prominent foci. Mathematical modeling and simulations incorporating the self-organization properties of Plk4 successfully generated the experimentally observed patterns. We therefore propose that the self-patterning of Plk4 is crucial for the regulation of centriole duplication. These results, defining the mechanisms of self-organized regulation, provide a fundamental principle for understanding centriole duplication.