Recent deep networks that directly handle points in a point set, e.g., PointNet, have been state-of-the-art for supervised learning tasks on point clouds such as classification and segmentation. In this work, a novel end-toend deep auto-encoder is proposed to address unsupervised learning challenges on point clouds. On the encoder side, a graph-based enhancement is enforced to promote local structures on top of PointNet. Then, a novel folding-based decoder deforms a canonical 2D grid onto the underlying 3D object surface of a point cloud, achieving low reconstruction errors even for objects with delicate structures. The proposed decoder only uses about 7% parameters of a decoder with fully-connected neural networks, yet leads to a more discriminative representation that achieves higher linear SVM classification accuracy than the benchmark. In addition, the proposed decoder structure is shown, in theory, to be a generic architecture that is able to reconstruct an arbitrary point cloud from a 2D grid. Our code is available at http://www.merl.com/research/ license#FoldingNet

Unlike on images, semantic learning on 3D point clouds using a deep network is challenging due to the naturally unordered data structure. Among existing works, Point-Net has achieved promising results by directly learning on point sets. However, it does not take full advantage of a point's local neighborhood that contains fine-grained structural information which turns out to be helpful towards better semantic learning. In this regard, we present two new operations to improve PointNet with a more efficient exploitation of local structures. The first one focuses on local 3D geometric structures. In analogy to a convolution kernel for images, we define a point-set kernel as a set of learnable 3D points that jointly respond to a set of neighboring data points according to their geometric affinities measured by kernel correlation, adapted from a similar technique for point cloud registration. The second one exploits local high-dimensional feature structures by recursive feature aggregation on a nearest-neighbor-graph computed from 3D positions. Experiments show that our network can efficiently capture local information and robustly achieve better performances on major datasets. Our code is available at http://www.merl.com/research/ license#KCNet

Summary Adenosine-to-inosine RNA editing is crucial for generating molecular diversity, and serves to regulate protein function through recoding of genomic information. Here, we discover editing within CaV1.3 Ca2+ channels, renown for low-voltage Ca2+-influx and neuronal pacemaking. Significantly, editing occurs within the channel’s IQ domain, a calmodulin-binding site mediating inhibitory Ca2+-feedback (CDI) on channels. The editing turns out to require RNA adenosine deaminase ADAR2, whose variable activity could underlie a spatially diverse pattern of CaV1.3 editing seen across the brain. Edited CaV1.3 protein is detected both in brain tissue and within the surface membrane of primary neurons. Functionally, edited CaV1.3 channels exhibit strong reduction of CDI; in particular, neurons within the suprachiasmatic nucleus show diminished CDI, with higher frequencies of repetitive action-potential and calcium-spike activity, in wildtype versus ADAR2 knockout mice. Our study reveals a mechanism for fine-tuning CaV1.3 channel properties in CNS, which likely impacts a broad spectrum of neurobiological functions.

Background: Alternative splicing generates calcium channel splice variants with altered electrophysiological properties. Results: Exclusion of exons encoding the IQb domain or proximal/distal domains attenuates Ca 2ϩ -dependent inactivation of the Ca V 1.3 channels. Conclusion: Alternative splicing at the C terminus alters the critical Ca 2ϩ inhibitory feedback property of Ca V 1.3 channels. Significance: Alternative splicing is an exquisite mechanism for customizing channel function within diverse biological niches.

Native Ca V 1.3 channels within cochlear hair cells exhibit a surprising lack of Ca 2ϩ -dependent inactivation (CDI), given that heterologously expressed Ca V 1.3 channels show marked CDI. To determine whether alternative splicing at the C terminus of the Ca V 1.3 gene may produce a hair cell splice variant with weak CDI, we transcript-scanned mRNA obtained from rat cochlea. We found that the alternate use of exon 41 acceptor sites generated a splice variant that lost the calmodulin-binding IQ motif of the C terminus. These Ca V 1.3 IQ⌬ ("IQ deleted") channels exhibited a lack of CDI, which was independent of the type of coexpressed ␤-subunits. Ca V 1.3 IQ⌬ channel immunoreactivity was preferentially localized to cochlear outer hair cells (OHCs), whereas that of Ca V 1.3 IQfull channels (IQ-possessing) labeled inner hair cells (IHCs). The preferential expression of Ca V 1.3 IQ⌬ within OHCs suggests that these channels may play a role in processes such as electromotility or activity-dependent gene transcription rather than neurotransmitter release, which is performed predominantly by IHCs in the cochlea.