Database: The Journal of Biological Databases and Curation最新文献

Curated resources at centralized repositories provide high-value service to users by enhancing data veracity. Curation, however, comes with a cost, as it requires dedicated time and effort from personnel with deep domain knowledge. In this paper, we investigate the performance of a large language model (LLM), specifically generative pre-trained transformer (GPT)-3.5 and GPT-4, in extracting and presenting data against a human curator. In order to accomplish this task, we used a small set of journal articles on wheat and barley genetics, focusing on traits, such as salinity tolerance and disease resistance, which are becoming more important. The 36 papers were then curated by a professional curator for the GrainGenes database (https://wheat.pw.usda.gov). In parallel, we developed a GPT-based retrieval-augmented generation question-answering system and compared how GPT performed in answering questions about traits and quantitative trait loci (QTLs). Our findings show that on average GPT-4 correctly categorized manuscripts 97% of the time, correctly extracted 80% of traits, and 61% of marker-trait associations. Furthermore, we assessed the ability of a GPT-based DataFrame agent to filter and summarize curated wheat genetics data, showing the potential of human and computational curators working side-by-side. In one case study, our findings show that GPT-4 was able to retrieve up to 91% of disease related, human-curated QTLs across the whole genome, and up to 96% across a specific genomic region through prompt engineering. Also, we observed that across most tasks, GPT-4 consistently outperformed GPT-3.5 while generating less hallucinations, suggesting that improvements in LLM models will make generative artificial intelligence a much more accurate companion for curators in extracting information from scientific literature. Despite their limitations, LLMs demonstrated a potential to extract and present information to curators and users of biological databases, as long as users are aware of potential inaccuracies and the possibility of incomplete information extraction.

Curated resources at centralized repositories provide high-value service to users by enhancing data veracity. Curation, however, comes with a cost, as it requires dedicated time and effort from personnel with deep domain knowledge. In this paper, we investigate the performance of a large language model (LLM), specifically generative pre-trained transformer (GPT)-3.5 and GPT-4, in extracting and presenting data against a human curator. In order to accomplish this task, we used a small set of journal articles on wheat and barley genetics, focusing on traits, such as salinity tolerance and disease resistance, which are becoming more important. The 36 papers were then curated by a professional curator for the GrainGenes database (https://wheat.pw.usda.gov). In parallel, we developed a GPT-based retrieval-augmented generation question-answering system and compared how GPT performed in answering questions about traits and quantitative trait loci (QTLs). Our findings show that on average GPT-4 correctly categorized manuscripts 97% of the time, correctly extracted 80% of traits, and 61% of marker-trait associations. Furthermore, we assessed the ability of a GPT-based DataFrame agent to filter and summarize curated wheat genetics data, showing the potential of human and computational curators working side-by-side. In one case study, our findings show that GPT-4 was able to retrieve up to 91% of disease related, human-curated QTLs across the whole genome, and up to 96% across a specific genomic region through prompt engineering. Also, we observed that across most tasks, GPT-4 consistently outperformed GPT-3.5 while generating less hallucinations, suggesting that improvements in LLM models will make generative artificial intelligence a much more accurate companion for curators in extracting information from scientific literature. Despite their limitations, LLMs demonstrated a potential to extract and present information to curators and users of biological databases, as long as users are aware of potential inaccuracies and the possibility of incomplete information extraction.

With the continuous advancements in cancer immunotherapy, neoantigen-based therapies have demonstrated remarkable clinical efficacy. However, accurately predicting the immunogenicity of neoantigens remains a significant challenge. This is mainly due to two core factors: the scarcity of high-quality neoantigen datasets and the limited prediction accuracy of existing immunogenicity prediction tools. This study addressed these issues through several key steps. First, it collected and organized immunogenic neoantigen peptide data from publicly available literature and neoantigen databases. Second, it analyzed the data to identify key features influencing neoantigen immunogenicity prediction. Finally, it integrated existing prediction tools to create TumorAgDB1.0, a comprehensive tumor neoantigen database. TumorAgDB1.0 offers a user-friendly platform. Users can efficiently search for neoantigen data using parameters like amino acid sequence and peptide length. The platform also offers detailed information on the characteristics of neoantigens and tools for predicting tumor neoantigen immunogenicity. Additionally, the database includes a data download function, allowing researchers to easily access high-quality data to support the development and improvement of neoantigen immunogenicity prediction tools. In summary, TumorAgDB1.0 is a powerful tool for neoantigen screening and validation in tumor immunotherapy. It offers strong support to researchers. Database URL: https://tumoragdb.com.cn.

With the continuous advancements in cancer immunotherapy, neoantigen-based therapies have demonstrated remarkable clinical efficacy. However, accurately predicting the immunogenicity of neoantigens remains a significant challenge. This is mainly due to two core factors: the scarcity of high-quality neoantigen datasets and the limited prediction accuracy of existing immunogenicity prediction tools. This study addressed these issues through several key steps. First, it collected and organized immunogenic neoantigen peptide data from publicly available literature and neoantigen databases. Second, it analyzed the data to identify key features influencing neoantigen immunogenicity prediction. Finally, it integrated existing prediction tools to create TumorAgDB1.0, a comprehensive tumor neoantigen database. TumorAgDB1.0 offers a user-friendly platform. Users can efficiently search for neoantigen data using parameters like amino acid sequence and peptide length. The platform also offers detailed information on the characteristics of neoantigens and tools for predicting tumor neoantigen immunogenicity. Additionally, the database includes a data download function, allowing researchers to easily access high-quality data to support the development and improvement of neoantigen immunogenicity prediction tools. In summary, TumorAgDB1.0 is a powerful tool for neoantigen screening and validation in tumor immunotherapy. It offers strong support to researchers. Database URL: https://tumoragdb.com.cn.

This perspective article synthesizes current knowledge regarding what is known regarding biocuration as a career and the challenges facing the field. It draws on existing literature and ongoing qualitative research to discuss the nature of biocuration, biocurators' career trajectories, key challenges that biocurators face, and strategies for overcoming these. Overall, biocurators express a high degree of satisfaction with their work and see it as central to the wider biosciences. The central challenges that they face relate to the underfunding and under-recognition of this work, meaning that there is minimal stable funding for the field and that the work of human biocurators is often invisible to those who use curated resources. The article closes by critically discussing existing and potential strategies for responding to these challenges.

The unprecedented volume of big data being routinely generated for nonmodel crop species, coupled with advanced technology enabling the use of big data in breeding, gives further impetus for the need to have access to crop community databases, where all relevant data are curated and integrated. Funding for such databases is, however, insufficient and intermittent, resulting in the data being underutilized. While increased awareness of the importance of funding databases is important, it is practically necessary to find a more efficient way to build a community database. To meet the need for integrated database resources for various crop genomics, genetics, and breeding research communities, we have built five crop databases over the last decade using an open-source database platform and software. We describe the system and methods used for database construction, curation, and analysis protocols, and the data and tools that are available in these five crop databases. Database URL: The Genome Database for Rosaceae (GDR, www.rosaceae.org), the Genome Database for Vaccinium (GDV, www.vaccinium.org), the Citrus Genome Database (CGD, www.citrusgenomedb.org), the Pulse Crop Database (PCD, www.pulsedb.org), and CottonGen (www.cottongen.org).

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been circulating and adapting within the human population for >4 years. A large number of mutations have occurred in the viral genome, resulting in significant variants known as variants of concern (VOCs) and variants of interest (VOIs). The spike (S) protein harbors many of the characteristic mutations of VOCs and VOIs, and significant efforts have been made to explore functional effects of the mutations in the S protein, which can cause or contribute to viral infection, transmission, immune evasion, pathogenicity, and illness severity. However, the knowledge and understanding are dispersed throughout various publications, and there is a lack of a well-structured database for functional annotation that is based on manual curation. AnnCovDB is a database that provides manually curated functional annotations for mutations in the S protein of SARS-CoV-2. Mutations in the S protein carried by at least 8000 variants in the GISAID were chosen, and the mutations were then utilized as query keywords to search in the PubMed database. The searched publications revealed that 2093 annotation entities for 205 single mutations and 93 multiple mutations were manually curated. These entities were organized into multilevel hierarchical categories for user convenience. For example, one annotation entity of N501Y mutation was 'Infectious cycle➔Attachment➔ACE2 binding affinity➔Increase'. AnnCovDB can be used to query specific mutations and browse through function annotation entities. Database URL: https://AnnCovDB.app.bio-it.tech/.