
Frameworks for Brain Cell Type Definition, Ontology, and Nomenclature Workshop: Mapping of Cell Type Data
In this virtual workshop, we will explore best practices for mapping single cell data to reference cell atlases. Mapping single cell data to reference cell atlases is complicated by batch effects between datasets, systematic differences in biological conditions of the samples, limited availability of computational resources, and sharing restrictions on raw data. While our focus is on transcriptomically-defined cell types of the brain, applications in other tissues and in the use of other data modalities will also be considered.
Thursday, May 12, 2022
8am-2pm Pacific Time
Registration: general attendee
General attendees are invited to listen to the full workshop and pose questions to the speakers. Registration is free.
Data challenge registration closed May 2.
Defining and naming cell types is a key problem of contemporary neuroscience. While the problem has preoccupied neuroscientists for over a century, it now has significantly increased attention due to our ability to collect cellular level data in a high-throughput manner. International consortia involved in molecular brain cell classification, including the BRAIN Initiative Cell Census Network (BICCN), the Human Cell Atlas (HCA), and the Human Biomolecular Atlas Program (HuBMAP) are generating and classifying cell types in all organ systems in the human body at a rapid pace. Standards for classification, naming, and data mapping are essential to form a common language of cell types to make sense of these enormous datasets, and to promote understanding between multiple domains, projects, and scientists.
Essential to our understanding of cell types and their functions is to catalog measured transcriptomes of cells belonging to each cell type, and to present this information in the form of comprehensive reference single cell atlases. Much like how reference genomes provide a map for locating the origin of sequencing reads, reference cell atlases can be used to map query cells to potential cell types in the reference atlas in order to rapidly characterize and compare relevant cell phenotypes. Reference atlases ultimately help quantify transcriptional heterogeneity that arises as a result of natural variation, aging, environmental influences, and disease. Large single-cell atlases comprising millions of cells across tissues, organs, donors, developmental stages, and conditions are now being generated by consortia such as the Human Cell Atlas and BICCN to serve as references for smaller-scale studies.
The workshop will include presentations about key state-of-the-art methods and techniques, and a comparison of results obtained by different methods in a hands-on data mapping challenge. To explore these mapping techniques, workshop participants will be provided with data mapping “challenges” to be completed before the workshop. Creative solutions to these challenges will have the opportunity to present to a national audience at the June 6, 2022, BRAIN Initiative Cell Census Network (BICCN) consortium meeting and participate in a published article detailing the challenge results.
Agenda
Workshop agenda is subject to change.
Session |
Speakers |
Time: Eastern |
Time: Pacific |
Zoom webinar opens to participants |
10:55am |
7:55am |
|
Welcome and housekeeping |
Workshop organizers |
11:00-11:05am |
8:00-8:05am |
Introduction, workshop goals, and overview of data challenges |
Mike Hawrylycz with Jesse Gillis, Gerald Quon, Richard Scheuermann, and Uygar Sümbül |
11:05-11:20am |
8:05-8:20am |
Mapping cell types in a biological context |
Ed Lein, Allen Institute for Brain Science |
11:20-11:30am |
8:20-8:30am |
Invited talk: Geometric and Topological methods for dissection of cellular heterogeneity |
Smita Krishaswamy, Yale School of Medicine |
11:30am-12:00pm |
8:30-9:00am |
Data Challenge 1: Presentations and discussion |
Data challenge teams and workshop organizers |
12:00-12:45pm |
9:00-9:45am |
Invited talk: Data Integration and Transfer Learning for Merging Single Cell Transcriptomics Data | Kathryn Roeder, Carnegie Mellon University | 12:45-1:15pm | 9:45-10:15am |
Break |
1:15-1:45pm |
10:15-10:45am |
|
Data Challenge 2: Presentations and discussion |
Data challenge teams and workshop organizers |
1:45-2:30pm |
10:45-11:30am |
Invited talk: Cell type homologies in the mammalian brain |
Trygve Bakken, Allen Institute for Brain Science |
2:30-3:00pm |
11:30am-12:00pm |
Invited talk: Iterative single-cell multi-omic integration using online learning |
Joshua D. Welch, University of Michigan |
3:00-3:30pm |
12:00-12:30pm |
Cell type data mapping: Strategic discussion |
Workshop organizers, invited speakers, and data challenge participants |
3:30-4:30pm |
12:30-1:30pm |
Closing remarks |
Richard Scheuermann and Mike Hawrylycz |
4:30-5:00pm | 1:30-2:00pm |
Data challenges
A key part of this workshop will be transcriptomic data mapping comparison of state-of-the-art approaches. We have prepared query and reference data sets based on a recent survey of brain-derived cell types and a BARseq2 dataset that jointly measures transcriptomic signatures of marker genes and spatial locations of neurons across a whole mouse cortex. Participants are asked to match cell types between the query and reference datasets and to identify cell types based on the expression data alone in the BARseq2 dataset and the cell types will then be assessed through their spatial patterns in the held-out physical locations. Participants may choose to approach Challenge 1, Challenge 2, or both. Data Challenge registration closed May 2.
During the workshop:
-
Present and benchmark contemporary algorithms for accurately mapping single cell transcriptomic data to standard cell atlases and taxonomies.
-
Discuss approaches for intra-platform and cross-platform single cell transcriptomic data normalization.
-
Discuss the challenges with producing optimal cluster partitioning.
-
Review methods for multimodal association of single cell data including morphology and electrophysiology data.
-
Review methods for mapping spatial transcriptomics data to reference cell types defined from sc/snRNA-seq data.
-
Evaluate existing data mapping algorithms supplied by BICCN consortium members on various new datasets; plan a roadmap for development of improved data mapping algorithms and infrastructure.
-
Discuss results and best practices from the data atlas mapping challenge on key data sets from sc/snRNA-seq and spatial transcriptomics data.
Workshop organizers
-
Jesse Gillis, University of Toronto
-
Mike Hawrylycz, Allen Institute for Brain Science
-
Gerald Quon, University of California, Davis
-
Richard Scheuermann, J. Craig Venter Institute
-
Uygar Sümbül, Allen Institute for Brain Science
Workshop support
-
NIH BRAIN Initiative grant, RFA-MH-19-146, A Community Framework for Data-driven Brain Transcriptomic Cell Type Definition, Ontology, and Nomenclature, 1RF1MH123220-01
-
NIH BRAIN Initiative Brain Cell Data Center (BCDC), RFA-MH-17-215, A Community Resource for Single Cell Data in the Brain, 1U24MH114827-01
-
Allen Institute for Brain Science