Simpson, T. I. | bioRxiv Science

bioRxiv2026

The Shape of Biological Metadata: Measuring Repository Richness with Entity-Based NLP Metrics

Ensuring the availability and accessibility of research data is fundamental to advancing knowledge, as codified in the FAIR principles (Findable, Accessible, Interoperable, and Reusable). Accurate metadata documentation is indispensable for meeting these principles; however, entries in deposition databases often contain inadequate, repetitive, or incomplete descriptions. Much of this metadata is captured in free-text fields, motivating the need for scalable, repository-agnostic methods to quantify metadata richness. Here, we quantify free-text metadata richness across three repositories using Natural Language Processing (NLP) methods: BioDare2, an experimental circadian rhythm database; DataShare, a domain-agnostic University of Edinburgh database; and Image Data Resource (IDR), a public repository of biological image datasets from published studies. In general, repositories exhibit distinct distributions of word count and information density, consistent with differences in scope. Named-entity recognition and information-density metrics detected significant category-level differences in BioDare2 (species) and DataShare (communities), while identifying greater consistency in the more curated IDR. The most frequent entities reflected each repository's focus: circadian terminology in BioDare2, microscopy-related entities in IDR, and community-driven terms in DataShare. We develop a scalable framework utilising word counts, named-entity recognition, and entity-derived information density to assess metadata quality across repositories, offering a broadly applicable evaluation tool.

bioinformatics↗

bioRxiv2026

cadmus: a robust pipeline for scalable retrieval of full-text biomedical literature

cadmus is an open-source Python toolkit for automated retrieval and processing of full-text biomedical literature. It utilises programmatic access to PubMed, Crossref, Europe PMC, PMC, and publisher APIs, allowing users to construct large, domain-specific corpora with minimal manual intervention. cadmus parses PDF, HTML, XML, and plain text files, standardising them for downstream biomedical text mining. During the retrieval of a Developmental Disorders Corpus (204,043 publications), it achieved an 85.2% full-text retrieval rate with institutional subscriptions and 54.4% without. To test the fidelity of retrieved full-texts, we used ScispaCy to infer the similarity of paired documents from 44,264 open-access PubMed Central files and the files retrieved from cadmus, resulting in an average cosine similarity score of 0.98. Rarefaction analyses demonstrated that full-text corpora double the coverage of unique biomedical concepts over abstracts, resulting in better access to the depth of the biomedical information available. Availability and implementationcadmus is a freely available package for non-commercial research at https://github.com/biomedicalinformaticsgroup/cadmus and released under the MIT License.

bioinformatics↗

bioRxiv2018

Analysis of proteins in computational models of synaptic plasticity

The desire to explain how synaptic plasticity arises from interactions between ions, proteins and other signalling molecules has propelled the development of biophysical models of molecular pathways in hippocampal, striatal and cerebellar synapses. The experimental data underpinning such models is typically obtained from low-throughput, hypothesis-driven experiments. We used high-throughput proteomic data and bioinformatics datasets to assess the coverage of biophysical models.\n\nTo determine which molecules have been modelled, we surveyed biophysical models of synaptic plasticity, identifying which proteins are involved in each model. We were able to map 4.2% of previously reported synaptic proteins to entities in biophysical models. Linking the modelled protein list to Gene Ontology terms shows that modelled proteins are focused on functions such as calmodulin binding, cellular responses to glucagon stimulus, G-alpha signalling and DARPP-32 events.\n\nWe cross-linked the set of modelled proteins with sets of genes associated with common neurological diseases. We find some examples of disease-associated molecules that are well represented in models, such as voltage-dependent calcium channel family (CACNA1C), dopamine D1 receptor, and glutamate ionotropic NMDA type 2A and 2B receptors. Many other disease-associated genes have not been included in models of synaptic plasticity, for example catechol-O-methyltransferase (COMT) and MAO A. By incorporating pathway enrichment results, we identify LAMTOR, a gene uniquely associated with Schizophrenia, which is closely linked to the MAPK pathway found in some models.\n\nOur analysis provides a map of how molecular pathways underpinning neurological diseases relate to synaptic biophysical models that can in turn be used to explore how these molecular events might bridge scales into cellular processes and beyond. The map illustrates disease areas where biophysical models have good coverage as well as domain gaps that require significant further research.\n\nAuthor summaryThe 100 billion neurons in the human brain are connected by a billion trillion structures called synapses. Each synapse contains hundreds of different proteins. Some proteins sense the activity of the neurons connecting the synapse. Depending on what they sense, the proteins in the synapse are rearranged and new proteins are synthesised. This changes how strongly the synapse influences its target neuron, and underlies learning and memory. Scientists build computational models to reason about the complex interactions between proteins. Here we list the proteins that have been included in computational models to date. For good reasons, models do not always specify proteins precisely, so to make the list we had to translate the names used for proteins in models to gene names, which are used to identify proteins. Our translation could be used to label computational models in the future. We found that the list of modelled proteins contains only 4.2% of proteins associated with synapses, suggesting more proteins should be added to models. We used lists of genes associated with neurological diseases to suggest proteins to include in future models.

neuroscience↗