Increasing metadata coverage of SRA BioSample entries using deep learning–based named entity recognition

Abstract

High-quality metadata annotations for data hosted in large public repositories are essential for research reproducibility and for conducting fast, powerful and scalable meta-analyses. Currently, a majority of sequencing samples in the National Center for Biotechnology Information’s Sequence Read Archive (SRA) are missing metadata across several categories. In an effort to improve the metadata coverage of these samples, we leveraged almost 44 million attribute–value pairs from SRA BioSample to train a scalable, recurrent neural network that predicts missing metadata via named entity recognition (NER). The network was first trained to classify short text phrases according to 11 metadata categories and achieved an overall accuracy and area under the receiver operating characteristic curve of 85.2% and 0.977, respectively. We then applied our classifier to predict 11 metadata categories from the longer TITLE attribute of samples, evaluating performance on a set of samples withheld from model training. Prediction accuracies were high when extracting sample Genus/Species (94.85%), Condition/Disease (95.65%) and Strain (82.03%) from TITLEs, with lower accuracies and lack of predictions for other categories highlighting multiple issues with the current metadata annotations in BioSample. These results indicate the utility of recurrent neural networks for NER-based metadata prediction and the potential for models such as the one presented here to increase metadata coverage in BioSample while minimizing the need for manual curation. Database URL: https://github.com/cartercompbio/PredictMEE

A search-based geographic metadata curation pipeline to refine sequencing institution information and support public health

“…Many automated or semi-automated methods aim to standardize metadata by clustering or mapping to ontologies ( 19 ); however, collecting precise metadata about locations remains a laborious and error-prone process ( 2 , 4 ). In this study, SGMC employed a semi-automated technique that utilizes cloud and generative artificial intelligence approaches (i.e., ChatGPT) for the curation and update of geospatial metadata from the NCBI SRA.…”

“…However, like all large datasets, the SRA may contain errors and inaccuracies ( 3 ). These errors may arise from human mistakes (e.g., typographical errors, incorrect entries, missing information), technical issues (e.g., faulty data submission, transfer, or processing due to issues with the software, hardware, or network), and/or lack of standardization (e.g., expanded generation of user-defined properties and infrequent use of controlled vocabularies during data submission over time) ( 4 , 5 ). As a result, SRA users may encounter absent or erroneous data across categories, including missing fields from data sources in the cloud, unclear synonyms, spelling variants, and heterogeneous sample data specification ( 2 ).…”

Machine Learning Data Analysis Highlights the Role of Parasutterella and Alloprevotella in Autism Spectrum Disorders

“…Our reanalysis again demonstrates, if needed, the importance of sharing clinical data with public databases. As previously observed, too many clinical articles are still published without the relative sample codes or are deposited with incorrect labels [ 99 , 100 ] and private requests remain unanswered. This is not only a limit to the reproducibility of published results but, as demonstrated by this study and several other meta-analyses, new important biological knowledge can be produced with ML and AI (Artificial Intelligence) approaches.…”

Artificial Intelligence in Pediatrics: Learning to Walk Together

“…It has been shown that as the amount and quality of the data increase, the output and performance of the system increase as well. 58 , 59 High-resolution, large datasets are ideal for ML applications. The importance of the scale of the dataset can be seen in studies and AI-based models where an extremely vast amount of data has been used to train the model, such as ChatGPT where billions of data were used, and in the study done by Liang et al 31 where more than 100 million data points were used and achieved a remarkable outcome.…”