HuBMAP MCs Demo Tools to Place Scientists’ Data into Genomic or Spatial Contexts, Leverage Visible Human Database to Learn

HuBMAP MCs Demo Tools to Place Scientists’ Data into Genomic or Spatial Contexts, Leverage Visible Human Database to Learn

March 8, 2021

This HuBMAP SciTech Webinar featured two Mapping Components (MCs) that have created online portals allowing scientists to enter their own data in a context that allows comparison and analysis with their MCs’ data sets. These two MCs’ efforts focus on transcriptomic context for single-cell data and spatial context for tissue images, respectively, with the latter also offering a massive, open online course based on the Visible Human database. A recording of the presentations can be found here.

Azimuth Demo

Rahul Satija of the New York Genome Center demonstrated Azimuth, a web-based tool that allows researchers to rapidly enable mapping and annotation single-cell sequencing profiles for as many as 100,000 cells or more. The approach, he explained, is inspired by a similar paradigm from the human genome project: Initial assembly and building a genome map is time- and effort-intensive, requiring high-quality data. However once that is accomplished, mapping new data to the reference genome is relatively fast and easy. Given the reference maps that are generated for Azimuth by Satija’s HuBMAP MC, accurate and robust comparisons and annotations of single-cell expression patterns can be efficiently accomplished.

Satija demonstrated how a database of a few thousand cells can be mapped into Azimuth in less than a minute—mapping of 100,000 cells or more takes seven to eight minutes, he added. Once entered, the data from each cell is projected into the reference map, receiving an annotation based on that reference as well as a prediction score to assign the level of conference for that mapping. A graphical display allows identification of cell type at a hierarchical level chosen by the researcher, matching the scientist’s data for a given cell to the most similar reference cell type in terms of gene activity as reflected in which kinds of RNA transcripts have been copied from the DNA. The system automatically computes the best markers for a given cell type and allows for comparison of cell types across species and disease states. Metadata such as gender, patient COVID status and spatial location are supported, and scientists can download anything from reference-map coordinates to individual cell annotations to further analyze on their local computer.

Azimuth can be accessed here.

HuBMAP Reference Atlas: ASCT+B Tables & 3D Reference Library Registration UI & VHMOOC Katy Börner and Andreas Bueckle of Indiana University, Bloomington, presented their Registration User Interface (RUI) and HuBMAP Visible Human Massive Open Online Course (VHMOOC) online tools. The former allows scientists to register their tissue-slice images spatially within a reference organ as well as with genetic and proteomic biomarkers (lipidomic and metabolomic references will be available in the future). The latter is an online learning tool that offers videos, hands-on exercises and self-paced quizzes based on the Visible Human dataset on issues in single-cell analysis, CCF mapping techniques and the HuBMAP portal. Both were developed by IUB’s The Human Body Atlas: High-Resolution, Functional Mapping of Voxel, Vector, and Meta Datasets MC.

Bueckle demonstrated how the RUI allows researchers to register tissue samples in a simple 3D interface that allows them to drag the sample into the reference organ while ensuring proper placement compared with organ anatomy, including substructures. Using the Visible Human dataset as a reference, a researcher can place the sample into the organ and relevant anatomical substructures quickly and accurately, using simple xyz coordinate motions as well as rotational movements of either the sample or the organ. The RUI then can allow exploration in the context of other tissues and other users’ samples, listing all relevant tissue sections in the database and allowing results to be saved as a json file.

Börner showed how users can use the VHMOOC to explore an overview of HuBMAP; tissue data acquisition and analysis; biomolecular data harmonization; CCF ontology, 3D reference objects and user interfaces; portal design and usage; open consent data and its advantages; Ontologies 101; and anatomical structures, cell types and biomarkers (AST+B) tables. Learning modules include use of the RUI and an exploration user interface (EUI) as well as NLM3D, a library of license-free 3D-anatomical models derived from medical imaging data by the National Library of Medicine. The group plans more such modules in the future.

Users can find the RUI here, and a preprint of a user study on the system here. The VHMOOC can be accessed here.