Can follow here: Beyond the Cell Atlas: Live – Chan Zuckerberg Biohub
- Using single cell to understand molecular physiology
- mostly around development of lung
- also touched on alzheimers
-
from CZ biohub
-
one of the leads on the tabula muris project - a map of all cells in mice in more than 20 tissues
-
a lot of labs involved
-
pipeline:
- annotate cells separately for each cell ( I think they used 10x)
- PCA -> cluster -> DE -> annotate using known markers
- no mention of high likelihood of batch effects
- showed a big tSNE plot of all the data combined (no details about how this was done)
- a little of bit benchmarking on no of genes detected across tissues to compare 10x, microwell and FACS (FACS most sensitive???)
- found combinations of TFs are important for distinguishing cell types
-
‘accelerating science’:
- about an imaging technique called vDISCO
- lots of pretty 3d figures on whole mouse
Cell atlas -> cell circuit & tissue circuit Focus on tissue circuit Few examples
-
airways mapping disease genes to cells
- knowing cells help determine where disease genes
- found new cell types in mouse trachea
- (again) transcription factors important for determining new cell types
- cystic fibrosis gene expressed by rare cell types
- known risk factor genes can be used to under their function in context of other disease like asthma
-
ulcerative colitis
- use scRNAseq to look at cell composition changes
- DE between healthy_inflamed_not inflamed
- look at cell-cell interaction using change in proportion of cell A with expression of ligand cell B (similar idea up a bit later in Sarah Teichmann’s talk)
- can use variation of cell type expression to predict GWAS gene functions
-
melanoma example modelling immunotherapy resistance with scRNAseq
- I got a bit lost here
single cell reconstruction of the maternal and fatal interface
- balance immuno modulation and tissue clearance
- key questions:
- what is cellular composition
- how do maternal cells avoid fetal rejection and control EVT invasion
- experimental design 11 decidua (maternal), 6 peripheral blood, 5 placentas
- integration of 10x and smart seq
- umap look at cell types and how they separate
- genotype sc by wGS integrating with scRNA-seq
- use t-cell repertoire to find clonal expansion (not sure how they did this)
- cell-cell communication using receptor (in one cell) and ligand (in another cell) networks using cellphonedb
- not sure of the exact details here but idea is to look at how network changes between maternal and fetal
- single cell genomics in brain is 95% discovery
- provides a quantitative framework for looking at specialisation and conservation of the brain
- understand diversity of cell types in cortex
- mouse, 1679 cells
- hierarchical clustering found 49 cell types
- repeated again with 10 times number of cells, found two times number of cell types
- similar result found in single nuclei sequencing Equivalent high-resolution identification of neuronal cell types with single-nucleus and single-cell RNA-sequencing | bioRxiv
- building human cortical taxonomy using surgery samples for epilepsy patients
- results cluster with mouse (see: Conserved cell types with divergent features between human and mouse cortex | bioRxiv
- found most cell markers were non-coding RNAs
- broad consensus between brain cell types between human and mouse (used CCA to combine datasets) to find homologous cell types
-
Cell types and lineage from single cell transcriptomes
-
classical approach: denovo clustering
-
mouse brains atlas: understand architecture of nervous system
-
droplet scRNA-seq
- 500,000 cells
- male and female mouse
- whole brain (sectioned)
- 265 cell types
- explains a lot of brain development and new cell types (like astrocytes) were found
-
beyond cell atlas:
- from static to dynamic: RNA velocity (see RNA velocity of single cells | Nature)
- unspliced mRNA from naturally occurring polyA/T in intron encoded in genome
- they are nascent
- can use ratio of spliced/unspliced to estimate RNA velocity (a few assumptions are such as constant degradation but can different if there are splice variants)
- can see independent processes such as cell cycle / lineage etc from this
Other links velocyto.org and mousebrain.org
- talk about a new technique called ORCA (optimal reconstruction of chromatin architecture) for measuring 3d structure of the genome (
- gave example of how genes in drosphilia
- talked about seed network RFA for chi
- at least 3 PIs
- one must be computational
- open until 13-Nov-18
- not just gut based (that is a seperate grant)
- talked about choanoflagellates and using that to infer animal evolution
- molecular exploration of the brain at single cell resolution
- individual brains vary a lot
- we can at look this with new tech (drop seq)
- comprehensive brain cell atlas
- cell type specific disease characterisation
- new tech drives biology (new cell types in striatum of brain)
- cell type annotation requires support from multiple modalities (samples, species, measurement types)
- need more data integration methods to reconcile variation at different modalities
- LIGER: integrative NMF plus take maximal loadings to form a neighoring graph
- lots of examples showing LIGER is great
- Regulation of spatial and temporal gene expression in an animal germline | bioRxiv talked about this paper
- spatial reconstruction from scRNAseq directly using optimal transport theory without imaging
- take expression distance and match it to physical distance (not sure how physical distance, or boundary conditions are determined)
- looks like it does a pretty good job
allencell.org vis interface for cell imaging