Search bioRxiv⌕ Search

Biology subjects

Dow, M.

Publications and source records attributed to Dow, M..

2 recordsLinked to original sources
bioRxiv2018

Creating a scalable deep learning based Named Entity Recognition Model for biomedical textual data by repurposing BioSample free-text annotations

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 Informations (NCBIs) 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 (AUROC) 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 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. AvailabilityAll the analyses, environments, and Jupyter notebooks pertaining to this manuscript are available on Github: https://github.com/cartercompbio/PredictMEE.

bioinformatics↗
bioRxiv2018

Extracting allelic read counts from 250,000 human sequencing runs in Sequence Read Archive

The Sequence Read Archive (SRA) contains over one million publicly available sequencing runs from various studies using a variety of sequencing library strategies. These data inherently contain information about underlying genomic sequence variants which we exploit to extract allelic read counts on an unprecedented scale. We reprocessed over 250,000 human sequencing runs (>1000 TB data worth of raw sequence data) into a single unified dataset of allelic read counts for nearly 300,000 variants of biomedical relevance curated by NCBI dbSNP, where germline variants were detected in a median of 912 sequencing runs, and somatic variants were detected in a median of 4,876 sequencing runs, suggesting that this dataset facilitates identification of sequencing runs that harbor variants of interest. Allelic read counts obtained using a targeted alignment were very similar to read counts obtained from whole genome alignment. Analyzing allelic read count data for matched DNA and RNA samples from tumors, we find that RNA-seq can also recover variants identified by WXS, suggesting that reprocessed allelic read counts can support variant detection across different library strategies in SRA. This study provides a rich database of known human variants across SRA samples that can support future meta-analyses of human sequence variation.

bioinformatics↗