Analysis and semantic representation of cell phenotype characteristics

National Library Of Medicine

Investigators

Linked publications, trials & patents

Abstract

Advances in sample preparation and sequencing technologies are now allowing for the comprehensive analysis of whole genome structure (epigenomic) and expression (transcriptomic) characteristics in individual single cells. This is revealing the cellular complexity of healthy, diseased, and perturbed tissues at unprecedented granularity. While general and project-specific repositories to capture and disseminate the primary (e.g., SRA) and processed data (e.g., GEO, ArrayExpress, CZI CELLxGENE, NeMO) have been developed and are being populated with data from published and unpublished studies, the knowledge derived from the analysis and interpretation of these experiments is currently only available as free text in scientific publications, making their exploration challenging and labor intensive, severely limiting the impact of these studies funded by the NIH and other agencies. The goal of this project is to develop and apply computational, statistical, and artificial intelligence methods for the analysis of high throughput cellular phenotypic data that will contribute to the development of a cell phenotype knowledgebase resource (NLM Cell Knowledge Network) that will capture knowledge about cell phenotypes, expose this knowledge for search and analysis by the extramural and intramural research communities, and integrate the cell phenotype knowledge with other sources of knowledge about diseases and drugs, especially with other NCBI and NLM resources, for discovery of novel diagnostic biomarkers and therapeutic targets. Two streams of knowledge are being used. First, using standard single cell genomics data formats (e.g., h5ad cell-by-gene expression matrices) captured from existing data repositories (e.g., CZI CELLxGENE, NeMO) as input, standardized and validated analysis pipelines have been implemented to produce information about cell type-specific marker genes and differential expression patterns, linked with experiment metadata about species and specimen sample sources, disease states, and perturbation responses. Second, using open access peer-reviewed publications reporting results from single cell genomics experiments from PubMed Central (PMC) as input, AI-driven (LLM) natural language processing (NLP) pipelines are being used to extract information about cell type-specific marker genes and differential expression patterns and their associations with disease states and perturbation responses. In order to keep pace with the rate and scale of single cell genomics experiments being reported in the literature, these knowledge streams have been implemented as NextFlow workflow pipelines. In order to maximize the interoperability of the derived knowledge about cell phenotypes with other sources of knowledge about genes, pathways, diseases, and drugs from other NLM/NCBI production resources, the output of these knowledge generating pipelines are translated into standardized semantically-structured assertions of subject-predicate-object triple statements for storage using semantic web technologies (e.g., RDF, OWL) and graph database (e.g., ArangoDB) platforms. The elements in the assertions are standardized with unique identifiers to be compatible with their representation in existing NLM/NCBI production resources to enable integration. NLM Cell will be exposed to the research community for query, exploration, and comparative analysis through an open access user interface for query and knowledge exploration in support of a number of mechanistic, diagnostic, and therapeutic use cases.