Poster #101 - Montana Knight

Evaluating Current Single-Cell Annotation Methods in Pediatric Populations

Knight, Montana Department of Biomedical and Health Informatics, Children's Hospital of Philadelphia, Philadelphia, PA, USA Svirydava2, Maryia Division of Allergy and Immunology, Department of Pediatrics, The Children's Hospital of Philadelphia, Philadelphia, PA, USA Campos, Jose S. Division of Allergy and Immunology, Department of Pediatrics, The Children's Hospital of Philadelphia, Philadelphia, PA, USA Henrickson, Sarah E. Division of Allergy and Immunology, Department of Pediatrics, The Children's Hospital of Philadelphia, Philadelphia, PA, USA

Single-cell sequencing provides novel insights into the substantial heterogeneity of cell populations. Many strides have been made in the computational processing of single-cell data to aide in further understanding this cellular diversity. New methods and tools are continuously introduced to help with everything from initial read mapping to downstream analysis1-3. One essential step of the standard pipeline is identifying the types of cells in the dataset2,4. There are available tools that automatically annotate cell types for you, but annotations are often done manually2,4. In fact, manual annotations are typically encouraged, despite it being subjective and a potential bottleneck in the process2,5,6. Automatic annotation could save time and improve reproducibility between studies, but can be discrepancies in the results among the tools out there, leaving researchers with the burden of figuring out which method and reference dataset fits best for their data5,6. This is exemplified with pediatric populations, as many of the popular reference atlases are in adults, which may not necessarily translate to children7-9. Here, we evaluate current automated annotation tools by applying them to peripheral blood mononuclear cells (PBMCs) in a publicly available adult cohort, and a pediatric cohort from the Children's Hospital of Philadelphia. Methods were further evaluated against curated manual annotations. Finally, applicability of these automated tools in the pediatric immune landscape was assessed.