Manual Reannotation Research Articles

Automated Grasp Labeling and Detection Framework with Pixel-Level Precision

Due to the differences in shape, material, and color of objects, the detection of planar grasps by robots remains challenging. Traditional methods rely on discrete grasp configurations for annotation, ignoring many possible grasp configurations. This leads to poor network generalization, making it challenging to handle a diverse range of grasped objects. Manual reannotation to continuous labels can effectively address this issue but comes with a significant cost. Therefore, this paper proposes a Pixel-level Grasp framework. Firstly, APGLG can automatically generate pixel-level grasp labels and discrete labels into continuous labels, effectively increasing the information content of individual data and improving the network generalization performance. Then, we propose the Max-Grasp-Net, built on the Multi-axis Vision Transformer and Dynamic Convolution Decomposition, to construct a U-shaped Network structure. A grasp decoder is incorporated, and deep supervision is applied to enhance network generalization. On the Cornell dataset, we achieve the best results, a grasping accuracy of 99.55%, an average success rate of 97.92% in single-object grasping, and 95.83% in multi-object grasping. We verified the effectiveness of our label generation algorithms and network innovation through actual grasp experiments.

An adaptive confidence-based data revision framework for Document-level Relation Extraction

Noisy annotations have become a key issue limiting Document-level Relation Extraction (DocRE). Previous research explored the problem through manual re-annotation. However, the handcrafted strategy is of low efficiency, incurs high human costs and cannot be generalized to large-scale datasets. To address the problem, we construct a confidence-based Revision framework for DocRE (ReD), aiming to achieve high-quality automatic data revision. Specifically, we first introduce a denoising training module to recognize relational facts and prevent noisy annotations. Second, a confidence-based data revision module is equipped to perform adaptive data revision for long-tail distributed relational facts. After the data revision, we design an iterative training module to create a virtuous cycle, which transforms the revised data into useful training data to support further revision. By capitalizing on ReD, we propose ReD-DocRED, which consists of 101,873 revised annotated documents from DocRED. ReD-DocRED has introduced 57.1% new relational facts, and concurrently, models trained on ReD-DocRED have achieved significant improvements in F1 scores, ranging from 6.35 to 16.55. The experimental results demonstrate that ReD can achieve high-quality data revision and, to some extent, replace manual labeling.11The ReD-DocRED is available at https://github.com/jc4357/ReD-DocRED.

PICO to PICOS: Weak Supervision to Extend Datasets with New Labels.

Hand-labelling clinical corpora can be costly and inflexible, requiring re-annotation every time new classes need to be extracted. PICO (Participant, Intervention, Comparator, Outcome) information extraction can expedite conducting systematic reviews to answer clinical questions. However, PICO frequently extends to other entities such as Study type and design, trial context, and timeframe, requiring manual re-annotation of existing corpora. In this paper, we adapt Snorkel's weak supervision methodology to extend clinical corpora to new entities without extensive hand labelling. Specifically, we enrich the EBM-PICO corpus with new entities through an example of "Study type and design" extraction. Using weak supervision, we obtain programmatic labels on 4,081 EBM-PICO documents, achieving an F1-score of 85.02% on the test set.

S1000: a better taxonomic name corpus for biomedical information extraction.

The recognition of mentions of species names in text is a critically important task for biomedical text mining. While deep learning-based methods have made great advances in many named entity recognition tasks, results for species name recognition remain poor. We hypothesize that this is primarily due to the lack of appropriate corpora. We introduce the S1000 corpus, a comprehensive manual re-annotation and extension of the S800 corpus. We demonstrate that S1000 makes highly accurate recognition of species names possible (F-score =93.1%), both for deep learning and dictionary-based methods. All resources introduced in this study are available under open licenses from https://jensenlab.org/resources/s1000/. The webpage contains links to a Zenodo project and three GitHub repositories associated with the study.

Resources for Turkish dependency parsing: introducing the BOUN Treebank and the BoAT annotation tool

In this paper, we introduce the resources that we developed for Turkish dependency parsing, which include a novel manually annotated treebank (BOUN Treebank), along with the guidelines we adopted, and a new annotation tool (BoAT). The manual annotation process that we employed was shaped and implemented by a team of four linguists and five Natural Language Processing (NLP) specialists. Decisions regarding the annotation of the BOUN Treebank were made in line with the Universal Dependencies (UD) framework as well as our recent efforts for unifying the Turkish UD treebanks through manual re-annotation. To the best of our knowledge, the BOUN Treebank is the largest Turkish UD treebank. It contains a total of 9761 sentences from various topics including biographical texts, national newspapers, instructional texts, popular culture articles, and essays. In addition, we report the parsing results of a state-of-the-art dependency parser obtained over the BOUN Treebank as well as two other treebanks in Turkish. Our results demonstrate that the unification of the Turkish annotation scheme and the introduction of a more comprehensive treebank lead to improved performance with regards to dependency parsing.

Safety and accuracy of the Rhythmia mapping system in pediatric patients

A new mapping system (Rhythmia) using a 64 mini-electrode small basket array (Orion) was developed that enables rapid high-density mapping in a short time. However, there are few reports about the usefulness of this system in pediatric cases. The purpose of this study was to investigate the safety and accuracy of the Rhythmia system and Orion catheter in children. Catheter ablation was performed using the Rhythmia system and Orion catheter in 23 patients younger than 20 years (body weight >20 kg) without a past medical history of cardiac disease. Mapping time, number of mapping beats, and number of mapping electrodes were compared for left atrium, right atrium and right ventricular outflow tract. Twenty-three maps of the right atrium were acquired in 12.6 minutes (range 8.9-15.1), consisting of 709 beats (range 492-1163) and 7132 electrograms (range 4618-10,533). Twelve maps of the left atrium were acquired in 12.1 minutes (range 9.8-14.6), consisting of 565 beats (range 446-881) and 6412 electrograms (range 4912-11,402). There were no significant difference in mapping time, accepted beats, and electrograms between the 2 chambers. Manual annotation was needed in 53 of 293,185 electrograms (0.01%) due to far-field ventricular electrogram sensing and artifact. No adverse events occurred in any of the cases. The Orion catheter and Rhythmia mapping system were safe and accurate for mapping various arrhythmias in pediatric patients. Detailed geometry and high-resolution activation mapping were acquired without the need for manual reannotation.

A manually annotated Actinidia chinensis var. chinensis (kiwifruit) genome highlights the challenges associated with draft genomes and gene prediction in plants

BackgroundMost published genome sequences are drafts, and most are dominated by computational gene prediction. Draft genomes typically incorporate considerable sequence data that are not assigned to chromosomes, and predicted genes without quality confidence measures. The current Actinidia chinensis (kiwifruit) ‘Hongyang’ draft genome has 164 Mb of sequences unassigned to pseudo-chromosomes, and omissions have been identified in the gene models.ResultsA second genome of an A. chinensis (genotype Red5) was fully sequenced. This new sequence resulted in a 554.0 Mb assembly with all but 6 Mb assigned to pseudo-chromosomes. Pseudo-chromosomal comparisons showed a considerable number of translocation events have occurred following a whole genome duplication (WGD) event some consistent with centromeric Robertsonian-like translocations. RNA sequencing data from 12 tissues and ab initio analysis informed a genome-wide manual annotation, using the WebApollo tool. In total, 33,044 gene loci represented by 33,123 isoforms were identified, named and tagged for quality of evidential support. Of these 3114 (9.4%) were identical to a protein within ‘Hongyang’ The Kiwifruit Information Resource (KIR v2). Some proportion of the differences will be varietal polymorphisms. However, as most computationally predicted Red5 models required manual re-annotation this proportion is expected to be small. The quality of the new gene models was tested by fully sequencing 550 cloned ‘Hort16A’ cDNAs and comparing with the predicted protein models for Red5 and both the original ‘Hongyang’ assembly and the revised annotation from KIR v2. Only 48.9% and 63.5% of the cDNAs had a match with 90% identity or better to the original and revised ‘Hongyang’ annotation, respectively, compared with 90.9% to the Red5 models.ConclusionsOur study highlights the need to take a cautious approach to draft genomes and computationally predicted genes. Our use of the manual annotation tool WebApollo facilitated manual checking and correction of gene models enabling improvement of computational prediction. This utility was especially relevant for certain types of gene families such as the EXPANSIN like genes. Finally, this high quality gene set will supply the kiwifruit and general plant community with a new tool for genomics and other comparative analysis.

Insights into metabolic osmoadaptation of the ectoines-producer bacterium Chromohalobacter salexigens through a high-quality genome scale metabolic model

BackgroundThe halophilic bacterium Chromohalobacter salexigens is a natural producer of ectoines, compatible solutes with current and potential biotechnological applications. As production of ectoines is an osmoregulated process that draws away TCA intermediates, bacterial metabolism needs to be adapted to cope with salinity changes. To explore and use C. salexigens as cell factory for ectoine(s) production, a comprehensive knowledge at the systems level of its metabolism is essential. For this purpose, the construction of a robust and high-quality genome-based metabolic model of C. salexigens was approached.ResultsWe generated and validated a high quality genome-based C. salexigens metabolic model (iFP764). This comprised an exhaustive reconstruction process based on experimental information, analysis of genome sequence, manual re-annotation of metabolic genes, and in-depth refinement. The model included three compartments (periplasmic, cytoplasmic and external medium), and two salinity-specific biomass compositions, partially based on experimental results from C. salexigens. Using previous metabolic data as constraints, the metabolic model allowed us to simulate and analyse the metabolic osmoadaptation of C. salexigens under conditions for low and high production of ectoines. The iFP764 model was able to reproduce the major metabolic features of C. salexigens. Flux Balance Analysis (FBA) and Monte Carlo Random sampling analysis showed salinity-specific essential metabolic genes and different distribution of fluxes and variation in the patterns of correlation of reaction sets belonging to central C and N metabolism, in response to salinity. Some of them were related to bioenergetics or production of reducing equivalents, and probably related to demand for ectoines. Ectoines metabolic reactions were distributed according to its correlation in four modules. Interestingly, the four modules were independent both at low and high salinity conditions, as they did not correlate to each other, and they were not correlated with other subsystems.ConclusionsOur validated model is one of the most complete curated networks of halophilic bacteria. It is a powerful tool to simulate and explore C. salexigens metabolism at low and high salinity conditions, driving to low and high production of ectoines. In addition, it can be useful to optimize the metabolism of other halophilic bacteria for metabolite production.

Clinical assessment and comparison of annotation algorithms in high‐density mapping of regular atrial tachycardias

High-density automated mapping of regular atrial tachycardias (ATs) requires accurate assessment of local activation times (LATs). To evaluate high-density mapping of ATs and compare the accuracy of different automated LAT annotation algorithms. Fifteen patients underwent AT ablation guided by the automated ConfiDENSEۛ high-density mapping module (Carto 3 v4) allowing manual reannotation (edited maps). For each AT, unedited automated maps were reconstructed offline by three algorithms: maximum unipolar slope (LATSlope ), bipolar peak (LATPeak ), and a new hybrid annotation algorithm (LATHybrid ). Five blinded experts were asked to define the (1) tachycardia mechanism, (2) ablation target, and (3) level of difficulty of these unedited maps. Twenty-one ATs (cycle length 300 ± 46ms, activation points 955 ± 421) were successfully ablated using LATHybrid guided ablation with manual editing in a small number of points. At 6 months, 14 (93%) of the patients were free of AT recurrences. Unedited LATHybrid maps showed the highest accuracy in defining the tachycardia mechanism (LATHybrid : 49%vs. LATPeak : 27%vs. LATSlope : 28%, P<0.001) and ablation target (LATHybrid : 65%vs. LATPeak : 39%vs. LATSlope : 31%, P<0.001). Overall, LATHybrid -annotated maps were graded as "easier to interpret" by the experts (difficulty score 2.3 ± 0.9) versus LATPeak (2.8 ± 1) and LATSlope (3.2 ± 0.8) (P<0.001). Only 12% of the LATHybrid maps were annotated as uninterpretable compared to 31% of LATSlope and 45% of the LATPeak maps (P<0.001). Automated LATHybrid annotation allows better and easier recognition of the tachycardia mechanism compared to automated LATPeak and LATSlope algorithms, although fully automated mapping still requires further improvements.

Identification of molecular markers in pectoral fin to predict artificial maturation of female European eels (Anguilla anguilla)

The European eel is a critically endangered species that cannot be reproduced in captivity yet. Artificial maturation of female European eels can be achieved via a laborious and expensive procedure, including weekly injections with pituitary extracts for up to 6months. The success rate is highly variable and a minimally invasive method for early selection of responsive eels would prevent the unnecessary and lengthy treatment of non-responding individuals. Since sexual maturation of European eels is accompanied by morphological changes of the pectoral fin, we examined whether fin could be used to monitor the response to the hormone treatment. Farmed eels were subjected to weekly injections with pituitary extracts and representative groups were sampled at 0 and 14–18weeks of hormone treatment. Responders and non-responders were identified based on the gonado-somatic index. Transcriptomes of pectoral fin samples obtained at the start and end of the trial were mapped using Illumina RNAseq. Responders showed 384 and non-responders only 54 differentially expressed genes. Highly stringent selection based on minimum expression levels and fold-changes and a manual re-annotation round yielded 23 up-regulated and 21 down-regulated maturation marker genes. The up-regulated markers belong to five categories: proteases, skin/mucus structural proteins, steroid hormone signaling, tyrosine/dopamine metabolism and lipid metabolism. The down-regulated markers are either blood markers or lectin-related genes. In conclusion, pectoral fin transcriptomes are a rich source of indicator markers for monitoring hormone induced sexual maturation of female European eels. In addition, these markers provide important new insight into several fundamental processes in eel biology.

New Assembly, Reannotation and Analysis of the Entamoeba histolytica Genome Reveal New Genomic Features and Protein Content Information

BackgroundIn order to maintain genome information accurately and relevantly, original genome annotations need to be updated and evaluated regularly. Manual reannotation of genomes is important as it can significantly reduce the propagation of errors and consequently diminishes the time spent on mistaken research. For this reason, after five years from the initial submission of the Entamoeba histolytica draft genome publication, we have re-examined the original 23 Mb assembly and the annotation of the predicted genes.Principal FindingsThe evaluation of the genomic sequence led to the identification of more than one hundred artifactual tandem duplications that were eliminated by re-assembling the genome. The reannotation was done using a combination of manual and automated genome analysis. The new 20 Mb assembly contains 1,496 scaffolds and 8,201 predicted genes, of which 60% are identical to the initial annotation and the remaining 40% underwent structural changes. Functional classification of 60% of the genes was modified based on recent sequence comparisons and new experimental data. We have assigned putative function to 3,788 proteins (46% of the predicted proteome) based on the annotation of predicted gene families, and have identified 58 protein families of five or more members that share no homology with known proteins and thus could be entamoeba specific. Genome analysis also revealed new features such as the presence of segmental duplications of up to 16 kb flanked by inverted repeats, and the tight association of some gene families with transposable elements.SignificanceThis new genome annotation and analysis represents a more refined and accurate blueprint of the pathogen genome, and provides an upgraded tool as reference for the study of many important aspects of E. histolytica biology, such as genome evolution and pathogenesis.

Evaluation of a large-scale biomedical data annotation initiative

BackgroundThis study describes a large-scale manual re-annotation of data samples in the Gene Expression Omnibus (GEO), using variables and values derived from the National Cancer Institute thesaurus. A framework is described for creating an annotation scheme for various diseases that is flexible, comprehensive, and scalable. The annotation structure is evaluated by measuring coverage and agreement between annotators.ResultsThere were 12,500 samples annotated with approximately 30 variables, in each of six disease categories – breast cancer, colon cancer, inflammatory bowel disease (IBD), rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), and Type 1 diabetes mellitus (DM). The annotators provided excellent variable coverage, with known values for over 98% of three critical variables: disease state, tissue, and sample type. There was 89% strict inter-annotator agreement and 92% agreement when using semantic and partial similarity measures.ConclusionWe show that it is possible to perform manual re-annotation of a large repository in a reliable manner.

ArrayIDer: automated structural re-annotation pipeline for DNA microarrays

BackgroundSystems biology modeling from microarray data requires the most contemporary structural and functional array annotation. However, microarray annotations, especially for non-commercial, non-traditional biomedical model organisms, are often dated. In addition, most microarray analysis tools do not readily accept EST clone names, which are abundantly represented on arrays. Manual re-annotation of microarrays is impracticable and so we developed a computational re-annotation tool (ArrayIDer) to retrieve the most recent accession mapping files from public databases based on EST clone names or accessions and rapidly generate database accessions for entire microarrays.ResultsWe utilized the Fred Hutchinson Cancer Research Centre 13K chicken cDNA array – a widely-used non-commercial chicken microarray – to demonstrate the principle that ArrayIDer could markedly improve annotation. We structurally re-annotated 55% of the entire array. Moreover, we decreased non-chicken functional annotations by 2 fold. One beneficial consequence of our re-annotation was to identify 290 pseudogenes, of which 66 were previously incorrectly annotated.ConclusionArrayIDer allows rapid automated structural re-annotation of entire arrays and provides multiple accession types for use in subsequent functional analysis. This information is especially valuable for systems biology modeling in the non-traditional biomedical model organisms.

Refined annotation and assembly of the Tetrahymena thermophila genome sequence through EST analysis, comparative genomic hybridization, and targeted gap closure

BackgroundTetrahymena thermophila, a widely studied model for cellular and molecular biology, is a binucleated single-celled organism with a germline micronucleus (MIC) and somatic macronucleus (MAC). The recent draft MAC genome assembly revealed low sequence repetitiveness, a result of the epigenetic removal of invasive DNA elements found only in the MIC genome. Such low repetitiveness makes complete closure of the MAC genome a feasible goal, which to achieve would require standard closure methods as well as removal of minor MIC contamination of the MAC genome assembly. Highly accurate preliminary annotation of Tetrahymena's coding potential was hindered by the lack of both comparative genomic sequence information from close relatives and significant amounts of cDNA evidence, thus limiting the value of the genomic information and also leaving unanswered certain questions, such as the frequency of alternative splicing.ResultsWe addressed the problem of MIC contamination using comparative genomic hybridization with purified MIC and MAC DNA probes against a whole genome oligonucleotide microarray, allowing the identification of 763 genome scaffolds likely to contain MIC-limited DNA sequences. We also employed standard genome closure methods to essentially finish over 60% of the MAC genome. For the improvement of annotation, we have sequenced and analyzed over 60,000 verified EST reads from a variety of cellular growth and development conditions. Using this EST evidence, a combination of automated and manual reannotation efforts led to updates that affect 16% of the current protein-coding gene models. By comparing EST abundance, many genes showing apparent differential expression between these conditions were identified. Rare instances of alternative splicing and uses of the non-standard amino acid selenocysteine were also identified.ConclusionWe report here significant progress in genome closure and reannotation of Tetrahymena thermophila. Our experience to date suggests that complete closure of the MAC genome is attainable. Using the new EST evidence, automated and manual curation has resulted in substantial improvements to the over 24,000 gene models, which will be valuable to researchers studying this model organism as well as for comparative genomics purposes.

Genome-wide analysis, evolutionary expansion, and expression of early auxin-responsive SAUR gene family in rice ( Oryza sativa)

Small auxin-up RNAs ( SAURs) are the early auxin-responsive genes represented by a large multigene family in plants. Here, we report the identification of 58 OsSAUR gene family members from rice ( Oryza sativa japonica cv Nipponbare), the model monocot plant, by a reiterative database search and manual reannotation; 2 of these are pseudogenes. The coding sequences of OsSAURs do not possess any intron. Most of the predicted OsSAUR protein sequences harbor a putative nuclear localization signal at their N-terminus. Localized gene duplications appear to be the primary genetic event responsible for SAUR gene family expansion in rice. Interestingly, the duplication of OsSAURs was found to be associated with the chromosomal block duplication as well. The phylogenetic analysis revealed that the SAUR gene family expanded in rice and Arabidopsis due to species-specific expansion of the family in monocots and dicots. The auxin-responsive elements and downstream element are conserved in the upstream and downstream sequences, respectively, of OsSAURs. In addition to the 21 OsSAURs with full-length cDNA sequences and 20 with expressed sequence tags, gene expression analyses of at least 7 OsSAURs by RT-qPCR indicated that the majority of identified OsSAURs most likely are expressed in rice. The transcript abundance of the OsSAURs examined increased within a few minutes of exogenous auxin application with varying kinetics. The present study provides basic genomic information for the rice SAUR gene family and will pave the way for deciphering the precise role of SAURs in plant growth and development.

Evolutionary Expansion, Gene Structure, and Expression of the Rice Wall-Associated Kinase Gene Family

The wall-associated kinase (WAK) gene family, one of the receptor-like kinase (RLK) gene families in plants, plays important roles in cell expansion, pathogen resistance, and heavy-metal stress tolerance in Arabidopsis (Arabidopsis thaliana). Through a reiterative database search and manual reannotation, we identified 125 OsWAK gene family members from rice (Oryza sativa) japonica cv Nipponbare; 37 (approximately 30%) OsWAKs were corrected/reannotated from earlier automated annotations. Of the 125 OsWAKs, 67 are receptor-like kinases, 28 receptor-like cytoplasmic kinases, 13 receptor-like proteins, 12 short genes, and five pseudogenes. The two-intron gene structure of the Arabidopsis WAK/WAK-Likes is generally conserved in OsWAKs; however, extra/missed introns were observed in some OsWAKs either in extracellular regions or in protein kinase domains. In addition to the 38 OsWAKs with full-length cDNA sequences and the 11 with rice expressed sequence tag sequences, gene expression analyses, using tiling-microarray analysis of the 20 OsWAKs on chromosome 10 and reverse transcription-PCR analysis for five OsWAKs, indicate that the majority of identified OsWAKs are likely expressed in rice. Phylogenetic analyses of OsWAKs, Arabidopsis WAK/WAK-Likes, and barley (Hordeum vulgare) HvWAKs show that the OsWAK gene family expanded in the rice genome due to lineage-specific expansion of the family in monocots. Localized gene duplications appear to be the primary genetic event in OsWAK gene family expansion and the 125 OsWAKs, present on all 12 chromosomes, are mostly clustered.

Functional genomic analysis of Arabidopsis thaliana glycoside hydrolase family 1

In plants, Glycoside Hydrolase (GH) Family 1 beta -glycosidases are believed to play important roles in many diverse processes including chemical defense against herbivory, lignification, hydrolysis of cell wall-derived oligosaccharides during germination, and control of active phytohormone levels. Completion of the Arabidopsis thaliana genome sequencing project has enabled us, for the first time, to determine the total number of Family 1 members in a higher plant. Reiterative database searches revealed a multigene family of 48 members that includes eight probable pseudogenes. Manual reannotation and analysis of the entire family were undertaken to rectify existing misannotations and identify phylogenetic relationships among family members. Forty-seven members (designated BGLU1 through BGLU47 ) share a common evolutionary origin and were subdivided into approximately 10 subfamilies based on phylogenetic analysis and consideration of intron-exon organizations. The forty-eighth member of this family ( At3g06510; sfr2 ) is a beta -glucosidase-like gene that belongs to a distinct lineage. Information pertaining to expression patterns and potential functions of Arabidopsis GH Family 1 members is presented. To determine the biological function of all family members, we intend to investigate the substrate specificity of each mature hydrolase after its heterologous expression in the Pichia pastoris expression system. To test the validity of this approach, the BGLU44 -encoded hydrolase was expressed in P. pastoris and purified to homogeneity. When tested against a wide range of natural and synthetic substrates, this enzyme showed a preference for beta -mannosides including 1,4- beta -D-mannooligosaccharides, suggesting that it may be involved in A. thaliana in degradation of mannans, galactomannans, or glucogalactomannans. Supporting this notion, BGLU44 shared high sequence identity and similar gene organization with tomato endosperm beta -mannosidase and barley seed beta -glucosidase/ beta -mannosidase BGQ60.