The mammalian prefrontal cortex comprises a set of highly specialized brain areas containing billions of cells and serves as the centre of the highest-order cognitive functions, such as memory, cognitive ability, decision-making and social behaviour. Although neural circuits are formed in the late stages of human embryonic development and even after birth, diverse classes of functional cells are generated and migrate to the appropriate locations earlier in development. Dysfunction of the prefrontal cortex contributes to cognitive deficits and the majority of neurodevelopmental disorders; there is therefore a need for detailed knowledge of the development of the prefrontal cortex. However, it is still difficult to identify cell types in the developing human prefrontal cortex and to distinguish their developmental features. Here we analyse more than 2,300 single cells in the developing human prefrontal cortex from gestational weeks 8 to 26 using RNA sequencing. We identify 35 subtypes of cells in six main classes and trace the developmental trajectories of these cells. Detailed analysis of neural progenitor cells highlights new marker genes and unique developmental features of intermediate progenitor cells. We also map the timeline of neurogenesis of excitatory neurons in the prefrontal cortex and detect the presence of interneuron progenitors in early developing prefrontal cortex. Moreover, we reveal the intrinsic development-dependent signals that regulate neuron generation and circuit formation using single-cell transcriptomic data analysis. Our screening and characterization approach provides a blueprint for understanding the development of the human prefrontal cortex in the early and mid-gestational stages in order to systematically dissect the cellular basis and molecular regulation of prefrontal cortex function in humans.

Human brain evolution is associated with expansion and folding of the neocortex. Increased diversity in neural progenitor (NP) populations (such as basally located radial glia [RG], which reside in an enlarged outer subventricular zone [OSVZ]) likely contributes to this evolutionary expansion, although their characteristics and relative contributions are only partially understood. Through single-cell transcriptional profiling of sorted human NP subpopulations, we identified the primate-specific TMEM14B gene as a marker of basal RG. Expression of TMEM14B in embryonic NPs induces cortical thickening and gyrification in postnatal mice. This is accompanied by SVZ expansion, the appearance of outer RG-like cells, and the proliferation of multiple NP subsets, with proportional increases in all cortical layers and normal lamination. TMEM14B drives NP proliferation by increasing the phosphorylation and nuclear translocation of IQGAP1, which in turn promotes G1/S cell cycle transitions. These data show that a single primate-specific gene can drive neurodevelopmental changes that contribute to brain evolution.

Conventional Zero-shot Learning approaches often suffer from severe performance degradation in the Generalised Zero-shot Learning (GZSL) scenario, i.e. to recognise test images that are from both seen and unseen classes. This paper studies the Class-level Over-fitting (CO) and empirically shows its effects to GZSL. We then address ZSL as a Triple Verification problem and propose a unified optimisation of regression and compatibility functions, i.e. two main streams of existing ZSL approaches. The complementary losses mutually regularise the same model to mitigate the CO problem. Furthermore, we implement a deep extension paradigm to linear models and significantly outperforms state-of-the-art methods in both GZSL and ZSL scenarios on the four standard benchmarks.