Evidence accumulated over the past decade shows that long non-coding RNAs (lncRNAs) are widely expressed and have key roles in gene regulation. Recent studies have begun to unravel how the biogenesis of lncRNAs is distinct from that of mRNAs and is linked with their specific subcellular localizations and functions. Depending on their localization and their specific interactions with DNA, RNA and proteins, lncRNAs can modulate chromatin function, regulate the assembly and function of membraneless nuclear bodies, alter the stability and translation of cytoplasmic mRNAs and interfere with signalling pathways. Many of these functions ultimately affect gene expression in diverse biological and physiopathological contexts, such as in neuronal disorders, immune responses and cancer. Tissue-specific and condition-specific expression patterns suggest that lncRNAs are potential biomarkers and provide a rationale to target them clinically. In this Review, we discuss the mechanisms of lncRNA biogenesis, localization and functions in transcriptional, post-transcriptional and other modes of gene regulation, and their potential therapeutic applications.

3D object detection is receiving increasing attention from both industry and academia thanks to its wide applications in various fields. In this paper, we propose the Point-Voxel Region based Convolution Neural Networks (PV-RCNNs) for accurate 3D detection from point clouds. First, we propose a novel 3D object detector, PV-RCNN-v1, which employs the voxel-to-keypoint scene encoding and keypoint-to-grid RoI feature abstraction two novel steps. These two steps deeply incorporate both 3D voxel CNN and PointNet-based set abstraction for learning discriminative point-cloud features. Second, we propose a more advanced framework, PV-RCNN-v2, for more efficient and accurate 3D detection. It consists of two major improvements, where the first one is the sectorized proposal-centric strategy for efficiently producing more representative and uniformly distributed keypoints, and the second one is the VectorPool aggregation to replace set abstraction for better aggregating local point-cloud features with much less resource consumption. With these two major modifications, our PV-RCNN-v2 runs more than twice as fast as the v1 version while still achieving better performance on the large-scale Waymo Open Dataset with 150m × 150m detection range. Extensive experiments demonstrate that our proposed PV-RCNNs significantly outperform previous state-of-the-art 3D detection methods on both the Waymo Open Dataset and the highly-competitive KITTI benchmark.

Atrophy patterns on MRI can reliably predict three neuropathological subtypes of Alzheimer’s disease (AD): typical, limbic-predominant, or hippocampal-sparing. A method to enable their investigation in the clinical routine is still lacking. We aimed to (1) validate the combined use of visual rating scales for identification of AD subtypes; (2) characterise these subtypes at baseline and over two years; and (3) investigate how atrophy patterns and non-memory cognitive domains contribute to memory impairment. AD patients were classified as either typical AD (n = 100), limbic-predominant (n = 33), or hippocampal-sparing (n = 35) by using the Scheltens’ scale for medial temporal lobe atrophy (MTA), the Koedam’s scale for posterior atrophy (PA), and the Pasquier’s global cortical atrophy scale for frontal atrophy (GCA-F). A fourth group with no atrophy was also identified (n = 30). 230 healthy controls were also included. There was great overlap among subtypes in demographic, clinical, and cognitive variables. Memory performance was more dependent on non-memory cognitive functions in hippocampal-sparing and the no atrophy group. Hippocampal-sparing and the no atrophy group showed less aggressive disease progression. Visual rating scales can be used to identify distinct AD subtypes. Recognizing AD heterogeneity is important and visual rating scales may facilitate investigation of AD heterogeneity in clinical routine.