Large-Scale Object Detection in the Wild With Imbalanced Data Distribution, and Multi-Labels.

Training with more data has always been the most stable and effective way of improving performance in the deep learning era. The Open Images dataset, the largest object detection dataset, presents significant opportunities and challenges for general and sophisticated scenarios. However, its semi-automatic collection and labeling process, designed to manage the huge data scale, leads to label-related problems, including explicit or implicit multiple labels per object and highly imbalanced label distribution. In this work, we quantitatively analyze the major problems in large-scale object detection and provide a detailed yet comprehensive demonstration of our solutions. First, we design a concurrent softmax to handle the multi-label problems in object detection and propose a soft-balance sampling method with a hybrid training scheduler to address the label imbalance. This approach yields a notable improvement of 3.34 points, achieving the best single-model performance with a mAP of 60.90% on the public object detection test set of Open Images. Then, we introduce a well-designed ensemble mechanism that substantially enhances the performance of the single model, achieving an overall mAP of 67.17%, which is 4.29 points higher than the best result from the Open Images public test 2018. Our result is published on https://www.kaggle.com/c/open-images-2019-object-detection/leaderboard.

GAIA-Universe: Everything is Super-Netify.

Pre-training on large-scale datasets has played an increasingly significant role in computer vision and natural language processing recently. However, as there exist numerous application scenarios that have distinctive demands such as certain latency constraints and specialized data distributions, it is prohibitively expensive to take advantage of large-scale pre-training for per-task requirements. we focus on two fundamental perception tasks (object detection and semantic segmentation) and present a complete and flexible system named GAIA-Universe(GAIA), which could automatically and efficiently give birth to customized solutions according to heterogeneous downstream needs through data union and super-net training. GAIA is capable of providing powerful pre-trained weights and searching models that conform to downstream demands such as hardware constraints, computation constraints, specified data domains, and telling relevant data for practitioners who have very few datapoints on their tasks. With GAIA, we achieve promising results on COCO, Objects365, Open Images, BDD100 k, and UODB which is a collection of datasets including KITTI, VOC, WiderFace, DOTA, Clipart, Comic, and more. Taking COCO as an example, GAIA is able to efficiently produce models covering a wide range of latency from 16 ms to 53 ms, and yields AP from 38.2 to 46.5 without whistles and bells. GAIA is released at https://github.com/GAIA-vision.

A CRISPR-Cas and Tat Peptide with Fluorescent RNA Aptamer System for Signal Amplification in RNA Imaging.

We reported on an efficient RNA imaging strategy based on a CRISPR-Cas and Tat peptide with a fluorescent RNA aptamer (TRAP-tag). Using modified CRISPR-Cas RNA hairpin binding proteins fused with a Tat peptide array that recruits modified RNA aptamers, this simple and sensitive strategy is capable of visualizing endogenous RNA in cells with high precision and efficiency. In addition, the modular design of the CRISPR-TRAP-tag facilitates the substitution of sgRNAs, RNA hairpin binding proteins, and aptamers in order to optimize imaging quality and live cell affinity. With CRISPR-TRAP-tag, exogenous GCN4, endogenous mRNA MUC4, and lncRNA SatIII were distinctly visualized in single live cells.

Live-cell RNA imaging using the CRISPR-dCas13 system with modified sgRNAs appended with fluorescent RNA aptamers.

The development of RNA imaging strategies in live cells is essential to improve our understanding of their role in various cellular functions. We report an efficient RNA imaging method based on the CRISPR-dPspCas13b system with fluorescent RNA aptamers in sgRNA (CasFAS) in live cells. Using modified sgRNA attached to fluorescent RNA aptamers that showed reduced background fluorescence, this approach provides a simple, sensitive way to image and track endogenous RNA with high accuracy and efficiency. In addition, color switching can be easily achieved by changing the fluorogenic dye analogues in living cells through user-friendly washing and restaining operations. CasFAS is compatible with orthogonal fluorescent aptamers, such as Broccoli and Pepper, enabling multiple colors RNA labeling or intracellular RNA-RNA interaction imaging. Finally, the visualization of severe fever with thrombocytopenia syndrome virus (SFTSV) was achieved by CasFAS, which may facilitate further studies on this virus.

Genome-wide sequencing and the clinical diagnosis of genetic disease: The CAUSES study.

Genome-wide sequencing (GWS) is a standard of care for diagnosis of suspected genetic disorders, but the proportion of patients found to have pathogenic or likely pathogenic variants ranges from less than 30% to more than 60% in reported studies. It has been suggested that the diagnostic rate can be improved by interpreting genomic variants in the context of each affected individual's full clinical picture and by regular follow-up and reinterpretation of GWS laboratory results. Trio exome sequencing was performed in 415 families and trio genome sequencing in 85 families in the CAUSES study. The variants observed were interpreted by a multidisciplinary team including laboratory geneticists, bioinformaticians, clinical geneticists, genetic counselors, pediatric subspecialists, and the referring physician, and independently by a clinical laboratory using standard American College of Medical Genetics and Genomics (ACMG) criteria. Individuals were followed for an average of 5.1 years after testing, with clinical reassessment and reinterpretation of the GWS results as necessary. The multidisciplinary team established a diagnosis of genetic disease in 43.0% of the families at the time of initial GWS interpretation, and longitudinal follow-up and reinterpretation of GWS results produced new diagnoses in 17.2% of families whose initial GWS interpretation was uninformative or uncertain. Reinterpretation also resulted in rescinding a diagnosis in four families (1.9%). Of the families studied, 33.6% had ACMG pathogenic or likely pathogenic variants related to the clinical indication. Close collaboration among clinical geneticists, genetic counselors, laboratory geneticists, bioinformaticians, and individuals' primary physicians, with ongoing follow-up, reanalysis, and reinterpretation over time, can improve the clinical value of GWS.

The design and rationale of the Advancing Cardiac Care Unit-based Rapid Assessment and Treatment of hypErcholesterolemia (ACCURATE) study.

Familial hypercholesterolemia (FH) is an inherited condition characterized by elevated low-density lipoprotein cholesterol (LDL-C) levels and premature atherosclerotic cardiovascular disease (ASCVD). Despite being the most common inherited cardiovascular disorder, it is still highly underdiagnosed and undertreated worldwide. We designed the Advancing Cardiac Care Unit-based Rapid Assessment and Treatment of hypErcholesterolemia (ACCURATE) study to test the hypothesis that opportunistic genetic testing for FH among patients hospitalized for acute coronary syndrome (ACS) will increase the diagnosis of FH and improve patient outcomes. ACCURATE is a non-randomized, controlled trial of patients <60 years old admitted to an acute cardiac unit with ACS and elevated LDL-C levels. The first cohort will consist of a control group of patients presenting with ACS who will be treated according to usual standard-of-care. The second cohort will consist of patients presenting with ACS in whom research-based genetic testing for FH will be performed during hospitalization and the results returned to the treating physicians. The primary endpoint will be the number of patients with a new diagnosis of FH. The secondary endpoints will be the proportion of patients who undergo intensification of lipid-lowering therapy, the lowest LDL-C level achieved, and the proportion of patients reaching guideline recommended lipid targets in the 12 months after the index ACS. To our knowledge, ACCURATE represents the first clinical trial of genetic testing for FH in the acute cardiac care setting and is expected to help identify optimal approaches to increase the diagnosis and treatment of FH.

Genome-wide sequencing as a first-tier screening test for short tandem repeat expansions.

BACKGROUND: Screening for short tandem repeat (STR) expansions in next-generation sequencing data can enable diagnosis, optimal clinical management/treatment, and accurate genetic counseling of patients with repeat expansion disorders. We aimed to develop an efficient computational workflow for reliable detection of STR expansions in next-generation sequencing data and demonstrate its clinical utility. METHODS: We characterized the performance of eight STR analysis methods (lobSTR, HipSTR, RepeatSeq, ExpansionHunter, TREDPARSE, GangSTR, STRetch, and exSTRa) on next-generation sequencing datasets of samples with known disease-causing full-mutation STR expansions and genomes simulated to harbor repeat expansions at selected loci and optimized their sensitivity. We then used a machine learning decision tree classifier to identify an optimal combination of methods for full-mutation detection. In Burrows-Wheeler Aligner (BWA)-aligned genomes, the ensemble approach of using ExpansionHunter, STRetch, and exSTRa performed the best (precision = 82%, recall = 100%, F1-score = 90%). We applied this pipeline to screen 301 families of children with suspected genetic disorders. RESULTS: We identified 10 individuals with full-mutations in the AR, ATXN1, ATXN8, DMPK, FXN, or HTT disease STR locus in the analyzed families. Additional candidates identified in our analysis include two probands with borderline ATXN2 expansions between the established repeat size range for reduced-penetrance and full-penetrance full-mutation and seven individuals with FMR1 CGG repeats in the intermediate/premutation repeat size range. In 67 probands with a prior negative clinical PCR test for the FMR1, FXN, or DMPK disease STR locus, or the spinocerebellar ataxia disease STR panel, our pipeline did not falsely identify aberrant expansion. We performed clinical PCR tests on seven (out of 10) full-mutation samples identified by our pipeline and confirmed the expansion status in all, showing absolute concordance between our bioinformatics and molecular findings. CONCLUSIONS: We have successfully demonstrated the application of a well-optimized bioinformatics pipeline that promotes the utility of genome-wide sequencing as a first-tier screening test to detect expansions of known disease STRs. Interrogating clinical next-generation sequencing data for pathogenic STR expansions using our ensemble pipeline can improve diagnostic yield and enhance clinical outcomes for patients with repeat expansion disorders.