Single Cell Genomics 2016

Notes from the Single Cell Genomics Conference, September 14-16 2016, Wellcome Genome Campus, Hixton, Cambridge, UK (Website)

Keynote Lecture: Genomic insights into human cortical development, lissencephaly, and Zika microcephaly; Arnold Kriegstein; University of California, San Francisco, USA

  • Single cell genomics to look in depth at developmental human brain
  • Clinical applications of identified cell types by genomic signature ie zika
  • Human brain is 1000x bigger than mice mostly in cortex
  • Elephants and proposes have bigger more folded brains but cell composition different
  • In mouse founder cells expand and move then divide: 13 cell cycles of neuromaturation
  • E13 comparable to GW11.5
  • As human mature have huge outer subventricular zone -> cell types in this area unclear
  • Morphological and location distinct but transcriptionally similar to radial glial and have similar markers -> call oRG
  • Distinctively jump when dividing; translocates along basal fiber
  • Smaller jump and fewer in mouse
  • In culture though just move in circles and divide
  • Single oRG can divide into lots of cell types and supports movement of other cell types with their fibers
  • Confirm diff exp marker genes for vRG and oRG using FISH (Pollen Cell paper 2015)
  • Build activated gene networks: gene set enrichment on marker genes(?) find oRG genes involved in stem mechanism for maintaining oSVZ
  • Old radial unit hypothesis from primates suggest one to one correspondence from one layer to another but data suggests only true at early stages, later stages have a discontinuous radial glial scaffold (unpublished) supragranular cortex expansion hypothesis
  • Hypothesize there are oRG like cells in glioblastoma; can see similar jumping and morphology
  • Use cerebral organoids (stem cells from adults to grow 3D brain) to model neurodevelopmental disorders like lissencephaly and microcephaly, neuronal maturation disorders
  • Early organoids between WT and lissencephaly derived look same and divides similarly in terms of vRG but abnormal in oRG -> cells jump further and don’t divide as readily in disease patient derived cells
  • Note can’t be modeled with mouse since mouse has smooth brains
  • Virus destroy brain and prvent new formation to cause microcephaly
  • Zika causes not just small but less folding in brain
  • Viral particles visible in infected tissue
  • Infected organoids are smaller since progenitor cells die
  • AXL necessary and sufficient receptor for infection of skin fibroblasts by zika
  • AXL highly enriched in oRG and blood vessels
  • Zika forms in clusters of radial glia in first trimester
  • Virus proteins go through entire cell even in the fibers
  • AXL antibody and knockdown block zika entry into astrocytes
  • Hypothesize AXL is mechanism of entry into brain
  • Anti AXL drugs are available but have side effects like blindness and sterility so not good for pregnant women; look at other potential treatments

Session 1: Neuroscience & Tissue Development Chair: Rickard Sandberg, Karolinska Institute, Sweden

Molecular census of mouse brain cell types; Sten Linnarsson; Karolinska Institute, Sweden

  • De novo cell type discovery
  • Cellular identity encoded in transcriptome; enough signal to make it work
  • Can visualize location of cells
  • Recently find neurons controlling goosebumps and nipple erection
  • Try to find oligodendrocyte lineage; arise from progenitors, divide into precursor then myelin forming then mature
  • Oligodendrocyte precursors and mature are stable state regardless of location in brain
  • Study developing midbrain at different development stages in both mouse and human; region of dopaminurgicn neuron development
  • Able to identify main neural cell types (known) by tsne
  • In mouse has fewer classes of progenitors possible due to gestation or age difference
  • High correspondence between mature cell types between mouse and human
  • Hypothesize progenitors divide less frequently in mouse ; timing is different
  • Use sequential smFish to look at radial glia in space across different developmental stages
  • Use gene markers To assess quality of created dopaminurgicn neurons for potential replacement therapy
  • Use regression to classify cells into known cell types to show high correspondence and ability to create all cell types

Reconstructing human organogenesis using single-cell RNAseq; Barbara Treutlein; Max Planck Institute for Evolutionary Anthropology, Germany

  • Paper: Treutlein et al Nature 2016
  • Start with somatic fibroblast and directly reprogram into different cell types without going through iPS (cell and tissue engineering) by over expression of transcription factors and certain genes
  • Analyze mef to induced neurons by scRNA-seq; reprogramming very inefficient only converts 10%
  • Cell cycle genes turn off and cell projection turns on in all cells
  • Induced neuron maturation has multiple cell fates (bifurcation) turns out some turn into myocytes and others turned appropriately into neurons
  • Decompose transitional states as weighted fractions of neuronal and fibroblast identifies (cell type signatures)
  • Along reprogramming cells go through NPC like state?
  • Use self organizing maps by grouping genes with correlated expression together
  • Get 2D expression picture for each cell then merged into group mean
  • Do for different populations to get all the over expression signatures in your data
  • Interesting radial signature plots and heat map like plots

Plasticity and heterogeneity of skin cells in health and tissue repair; Maria Kasper; Karolinska Institute, Sweden

Zeisel paper and something in press cell systems

  • Hair follical is easily accessible self maintaining mini organ; tool for studying stem cell, tomourgenesis, wound repair
  • One hair follicle has many stem cell populations all independent
  • Stem cell pops different because of location or intrinsic?
  • Stem cells migrate out of follicle during wound healing and get reprogrammed
  • Mutation in deep follicle cells causing small tumor but if wound induced then large tumors occur where wound is
  • Reconstruct IFE differentiation using spade trees
  • Identify point of no return in differentiation pathway Speculation:
  • Differentiation status influences stemness
  • Spatial determines function

Single nucleus RNA-Seq reveals dynamics of adult neurogenesis; Naomi Habib; Broad Institute, USA

  • Recent evidence of adult neural stem cells
  • In dentate gyrus niche within the hippocampus
  • To study neuron need enzymatic dissociation of cell body or just use single nuclei Scnucseq (Habib science 2016)
  • Use mice up to 2 years of age and can still get high quality sequencing
  • Label rare newborn cell throughout differentiation process -> div seq which just florescently label dividing cells
  • Order by maturation trajectory then cluster genes to show major transcriptional waves
  • Subset of immature neurons identified in spinal cord
  • BiSNE? Bi clustering on sne

Human cerebral organoids recapitulate gene expression programs of fetal neocortex development; Gray Camp; Max Planck Institute for Evolutionary Anthropology, Germany

  • Paper: Camp et al PNAS 2015
  • How well are genetic programs recapitulated by organoids?
  • Gene expressions are highly correlated
  • Networks derived from IPSC models also malformed like in disease patients so faithful recapitulation by model systems Surprising
  • Chimpanzee organoids recapitulate chimpanzee programs as well
  • Pc1 separated progenitors and neurons while Pc2 separated human and chimp but batch effects? So use scde :)
  • No obvious correlation though in terms of diff exp genes in human and mice between progenitor and neuron
  • Organoids have elongated metaphase in human vs chimp specific to apical progenitors (mouse even shorter) also orangutan as out group
  • Chimp organoids seems to mature faster than human though
  • Currently following up on specific genes

Characterization of adult neurogenic niches at single cell resolution; Marta Rodriguez Orejuela; Max Delbrueck Center for Molecular Medicine, Germany

  • Drop seq of dengate gyrus and svz
  • Mostly quality control stage proof of concept
  • 50% cells have more than 1000 genes ehhh
  • No difference between live and fixed cells?
  • Tsne identifies main types and lots of small groups?
  • Want to dcirbe clusters

Session 2: Chromatin Structure and Organisation Chair: Ido Amit, Weizmann Institute, Israel

Single cell dynamics of clonal memory; Amos Tanay; Weizmann Institute of Science, Israel

  • How transcriptional output is being dictated by intrceullar messages and combinations of transcription factors etc
  • Complexity introduced by epigenetics; composition of molecular organization introduces some time of epigenetic memory
  • Analyze 10000 singl cells from mouse embryo; cluster cells by select known markers
  • Cluster cells then cluster the clusters in relation to each other for more robust similarity
  • For each of clusters look for differentially enriched transcription factors
  • Over 100 specific TFs in the embryo
  • Correlation between fraction of TF specific to cluster coupled with faction of cluster specific gene expression
  • Ex primitive erythrocytes have low specific TF and low specific gene expression
  • Primitive erythrocytes may run yeast like promoter centric program
  • Suggests there are TF fields; gradients tht affect most cells often nested and overlapping not hierarchical
  • Now look for hypo methylation to detect enhancers nd qautnfiy fraction of cells with enhancer active on pop level
  • Multiple epigenetic enhancers form hubs, pair wise interactions to support epigenetic barrier?
  • Stability of enhancer landscape and interactions?mlots of enhancers are de novo ie only observed at this specific developmental time point

ATAC-ing regulatory variation in single cells; Will Greenleaf; Stanford University, USA

  • What are the are underpinnings of epigenetic landscape?
  • “Regulome”
  • Elements that are on are accessible while off are inaccessible
  • Atacseq get peaks where are accessible
  • Predict 8 class chromatin state using both length and position of reads
  • Apply to bulks of hematopoietic progenitor lineage
  • Many binary differences in epigenetic landscape
  • Clustering on atac gives cleaner grouping than gene expression since so binary
  • Study aml characterize diseases progenitors as mixes of normal pops suggesting use of many progenitor lineages ie lineage infidelity
  • Scatacseq allows for dividing cells by accessibility of 1 factor then aggregating to look at peaks in the two groups
  • Correction shows chromosome looping like hic data
  • Now look at single cell atacseq of hematopoietic differentiation
  • Use principal component of bulk to apply onto single cells for projection into low dimensional representation
  • Can see gradient of Gata and id3 motif accessibility
  • Tsne to group TFs to show which TFs tend to come together or have similar patterns of accessibility
  • Caution: naked dna contamination screws things up so background goes up with dead cells

Chromatin accessibility, DNA methylation and gene expression from the same single-cell; Stephen Clark; Babraham Institute, UK

  • Methylase enzyme label dna that’s exposed by methylation
  • Nome-seq
  • Now delve lop scmtseq; combine gtseq smartseq2 and bisulfiteseq
  • Separate gc and cg to get accessibility
  • Currently in proof of concept and development phase
  • Accessibility within gene correlated with higher expression as expected

Genome Architecture Mapping, new approach to map chromatin contacts; Ana Pombo; Max Delbrueck Center for Molecular Medicine, Germany

  • From imaging can see chromosome folding is not random but also not deterministic
  • Single point mutation on enhancer causes polydactyly
  • Many different mechanisms disrupt chromatic contacts causing many different phenotypes
  • What are he regulatory regions and target gees ? Real physical distance and frequency of contacts? How contacts established?
  • Address using technology called genome architecture mapping GAM
  • Sample using thin cryosections; cosegregation of nuclear profiles (how often colored together) implies distance
  • Requires many many nuclei? If you want to see 30kb windows then need 300 cryosections
  • Can identify tads so can be identified by non C tech
  • Some regions are contact and interacting because specific to function and others re bystanders;
  • look across population; probability interacting based on how often observed together in slice
  • Most contacts involve active genes and enhancers as expected
  • Get finer resolution by Identify which windows ar intacting and dividing window and assess frequency of cosegregation of finer windows
  • Multiple super enhancers contact simultaneously in embryonic stem cells

Chromosome dynamics revealed by single cell HiC; Peter Fraser; Babraham Institute, UK

  • Single cell Hi-C contacts suggest spherical chromosome arrangement
  • Laminar associated domains tend to be on outside
  • Diagonals (transmalignment) in single cell Hi-C suggest alignment during mitosis division so can see arrangement by spindles during division
  • Can infer cell cycle this way based on intachrompspmal contact decay
  • Chromosome paint shows genes point inwrd PhysicallY
  • Still very sparse coverage so can only make inference on what’s consistently observed as opposed to what’s not observed
  • Concern that hic is nuceosome rich and not capturing organization of all regions; may not be nice and spherical

Single-cell ATAC-seq identifies epigenetic differences in human pancreatic islet cell subtypes from normal and diabetic donors; Amanda Ackerman; Children’s Hospital of Philadelphia

  • Islet cells in pancreas impact diabetes, has known lineage- pancreas progenitor differentiate into alpha beta and exocrine cells
  • Alpha does glucagon secretion and beta does insulin secretion
  • In disease get dedifferentiation of beta cells or transdifferentiate into alpha cells
  • Plan is to apply to diabetic as well as obese but not diabetic patients
  • Possibly 4 subtypes of beta cells with different functions (speculative)

Dissecting Deregulated Enhancer Activity in Primary Leukemia Cells Jan-Philipp Mallm German Cancer Research Center (DKFZ), Germany

  • Use chipseq to look at his tone modifications
  • Differential regions of modification between normal and cll B cells
  • Can we deactivate lost enhancers?
  • Performs scatac sequencing of cll patients
  • Panobinostat treatment activate lost enhancers in cll
  • Extract enhancer promoter wiring to refine GO groups
  • Lmp1 virus transform cll cells through enhancers

Session 3: Immunology Chair: John Marioni, EMBL-EBI & CRUK CU, UK

Understanding Cellular Heterogeneity; Sarah Teichmann; Wellcome Trust Sanger Institute, UK

  • Sensitivity and specificity of scRNA-seq protocols
  • Compare smartseq, strtseq, marsseq, all techniques found
  • Based on Ercc spike ins (bioinformatics metadata analysis)
  • Endogenous rna may not be naked like spike ins so maybe more difficult to capture
  • so compare Ercc with smFISH genes
  • Show endogenous genes are captured at higher efficiency (countr intuitive)
  • Freeze thaw cycles decrease rna content by 20%
  • Finds high accuracy for all methods though expected lower for simultaneous methods like gtseq
  • Sensitivity drops more dramatically for some protocols (10^5-6 best)
  • Published in biorxiv

  • Reconstruct T cell differentiation by aligning cells by pseudo time using Gaussian process latent variable model
  • Encod information about time as a prior
  • Work or valentine scwennson who is interested in our lab
  • Identify bifurcation point: when cell decides fate to th1 and TFH
  • Fate bifurcation coincides with proliferation peak; look for trend correlations in different pseudo times
  • Extract genes associated with one fate or another
  • Use TCR to show that two different cell types can come from the same progeitor; TRACER software

Immunology in the age of single cell genomics; Ido Amit; Weizmann Institute of Science, Israel

  • Need new molecular microscope
  • Communication between immune and tissue cells dictate fate (matcowich science 2016) ; micro biome also impacts cell types (cell 2016)
  • Combine with crispr to confirm function of guide rna and their combinations

Long-term single cell quantification: New tools for old questions; Timm Schroeder; ETH Zurich, Switzerland

  • Need continuous single cell quantification to unambiguously resolve tree
  • Pseudo time ordering cannot recapitulate oscillations
  • Pseudo time can distort real time dynamics
  • TTT tracking softy analyze movies
  • Find cells before and after decision to try to find regulators but decision may be mad far in advance; use live quantification in model mice

Transcriptional heterogeneity and lineage commitment in hematopoietic progenitors; Amir Giladi; Weizmann Institute of Science, Israel

  • Finding differentiation manifolds
  • Like pca but plot diffusion components
  • Apply to visualize early blood development on scrtpqcr
  • Software: destiny in R

Single-cell RNA-seq-based identification and characterisation of somatic stem cells in adipose tissue & beyond; Bart Deplancke; Ecole Polytechnique Federale de Lausanne, Switzerland

  • Study fat biology due to increasing obesity and associated cancer and immunological disorders
  • Adipocytes differentiate from MSC, which are very diverse
  • Marker based analysis of preadipostem cells not sufficient so use unbiased tsne
  • Find subpopulation with paracrine signaling between cells that have refractory and inhibitory function (modulation) lack of this subpopulation may link to increased adipo genesis

  • ASAP- interactive implemetation of lots of algorithms with no installation
  • Cloud based
  • Unpublished will post in bioarxiv

  • Discoseq

Unbiased whole tissue analysis of the single cell transcriptional landscape of colon cancer; Matan Hofree; Broad Institute of MIT and Harvard, USA

  • Droplet based analysis of whole colon tumor
  • 10000 cells 1000 expressed genes

Diffusion pseudotime identifies lineage choice and graded transitions in myeloid progenitors; Fabian Theis; Helmholtz Center Munich, Germany

  • Paul et al cell 2015 finds marker genes for hematopoietic lineage suggests all stem cells are already primed
  • Do unbiased sequencing then mark by known markers to identify distinct populations , already committed
  • Focus on cells without strong lineage markers
  • Want connection with main cluster
  • But cluster relationship may be driven by cell cycle or stress
  • Find dormancy is coupled with increased expression of stemness genes
  • Hsc can be dormant or not
  • Propose new reference model for hematopoiesis with early erythrocytes differentiation

Session 4: Transcriptomics Chair: Sten Linnarsson, Karolinska Institute, Sweden

Single-cell gene expression analyses of allelic transcription and regulation; Rickard Sandberg; Karolinska Institute, Sweden

  • Paper: Petropoulos cell 2016
  • Models of allergic expression: genetic imprinting, X chromosome inactivation, allelic exclusion, fixed widespread autosomal random mono allelic expression
  • Use smartseq2
  • Use SNPs from crossed mice to distinguish alleles
  • Two conflicting theories of x inactivation: hyunh nature 2003 and okamoto nature 2005
  • Paternal/maternal X chromosome expression goes up then down suggesting de novo inactivation; coincides with xist activation
  • See chromosome wid pattern of lowered expression coinciding with xist expression; normalization/sampling bias?
  • Yet biallelic expression during dosage compensation; biallelic dampening as cells mature and xist get turned on
  • Different for mouse and human though

  • Paper: Reining nature genetics in press
  • Stochastic allelic fluctuations due to transcriptional bursting (sandberg nature genetics review 2015)
  • Yet Sasha claims inherited mono allelic expression
  • Therefore analyze random mono allelic gene expression of clones of primary fibroblasts
  • Find that among different clones mono allelic genes are not consistent
  • Mono allelic genes with high pvalue all lowly expressed(?) Unclear why pvalue so high if only based on 2 copies of gene
  • Haplotype phasing R package

Population balance reconstruction of differentiation hierarchies in developing and adult tissues by single cell droplet RNA-Seq; Allon Klein; Harvard Medical School, USA

  • 7 cents per cell including tech salary
  • Now getting 12000 umi or 4000 genes per cell
  • SPRING for manifold discovery: first pca then knn, keep edges in force directly graph, interact with visualization since low dimension can never capture all complexity

  • Predicting cell fate tool PBA
  • Model using population flux balance analysis
  • Solve in high dimension via nearest neighbor graph
  • Cell trajectories are inferred by potential field sinks
  • Predict fate probabilities
  • At every fate branch point assess diff exp to characterize

Early metazoan cell type evolution by single cell RNA-seq analysis; Arnau Sebé-Pedrós; Weizmann Institute of Science, Israel

  • Nematostellao whole organism scrnaseq
  • Only 200 umi per cell on average
  • Very high Cell size variability
  • Cluster using highly expressed genes
  • Poor annotation of 3’ utr are major issues for designing umis

Sequencing Small-RNA transcriptome of individual cells; Omid Faridani; Karolinska Institute, Sweden

  • Add adapter to both end of small rna
  • Issue is rRNA and lots of adaptor diners
  • For bulk can use size selection but can’t in single cell
  • In single cell eliminate rRNA, add 3’ adaptor, digest free adaptors, then add 5’ adaptor before cDNA synthesis
  • Small rna only 18 to 40 nt
  • Lots of small rna drives from mRNA; not just degraded
  • Serial dilution suggests 40% efficiency

Revealing novel cell types, cell-cell interactions, and cell lineages by single-cell sequencing; Alexander van Oudenaarden; Hubrecht Institute-KNAW, The Netherlands

  • Retaining spatial information via local neighborhoods and interactions
  • Manually separate cells physically attached together so now known at some point they were together
  • Do for 1000 single cell
  • Cluster and classify cells
  • In analysis connect cells if physically connects
  • Compar with random interactions
  • See which connections at more likely than random
  • Macrophages tend to be next to erythroblasts
  • B cells to to be next to early neutrophils
  • Megakaryocytes next to matur neutrophils

  • Paper: Mooijman et al nature biotech 2016
  • Use 5 hydroxymethylation to track lineage
  • Look at oocyte that divides
  • For one chromosome, strand is either methylated or not and sister chromosome is opposite
  • Use this to identify sister cells

On the widespread and critical impact of systematic bias and batch effects in single-cell RNA-Seq data; Stephanie Hicks; Harvard School of Public Health, USA

Seq-Well: A Portable Single-Cell RNA-Seq Platform for LowInput Clinical Samples; Marc Wadsworth; MIT, USA

  • Clinical complications: need scalability, portability, high capture efficiency
  • Uses microwell with membrane for sealing To confine mRNA t prevent cross contamination and increase efficiency
  • Comparable capture efficiency and library complexities to dropseq
  • Load beads night befor, 30 minut from cell loading to losing

Single cell preservation for RNAseq; Eshita Sharma; Wellcome Trust Centre for Human Genetics, UK

  • DSP a reversible cross linker to fix tissue sections shows high correspondence between fixed and fresh cells in bulk

Single-cell transcriptomics and functional analysis of single cells; Marc Unger; FLUIDIGM, UK

  • HT mRNA seq
  • $4500 for 800 cells for 6000 genes per cell
  • new 800 cell chip coming out soon, 10000 cell chip coming out next year

Dissection of T1 Diabetes progression using Single cell RNA sequencing of a RAT model; Manuel Garber; University of Massachusetts Medical School, USA

  • Study beta islet cells
  • Immune activation causes decline of beta cell resulting in abnormal blood sugar
  • Create rat model since generally type 1 diabetes resistant
  • Can induce t1D using pic+krv treatment (aggivamt)
  • Show same proportion of cell types as human
  • Do time course to map disease progression; islets taken from different days

Session 5: Imaging and Modelling Chair: Alexander van Oudenaarden, Hubrecht Institute, The Netherlands

In situ transcription profiling in tissues by seqFISH; Long Cai; California Institute of Technology, USA

  • Get past fluorescence overlap by sequential hybridization on fixed cells
  • Barcode corresponds to mRNA
  • Problem is with applying to thick tissue samples; high background
  • Now use hybridization chain reaction
  • So amplify 24 rounds after hybridization to amplify then digest and repeat with new probe
  • Apply to 125 genes in cortex and hippocampus
  • Design barcodes so that even with dropouts will be ok
  • Identify transcriptional clusters then map back on to image
  • See astrocytes ar spatially co localized but tanscriptionally distinct

Dynamic and heterogeneous DNA methylation in pluripotent cells; Heather Lee; Babraham Institute, UK

  • ESC very homogenous when naïve but hetrogeneeous when primed
  • What is the relationship between DNA methylation heterogeneity and phenotype
  • Variance of Dna methylation is greatest at enhancers
  • Naïve esc have low Dna methylation but primed esc vary
  • Turnover of cytosine modifications generate Dna methylation heterogeneity
  • DNA methylation heterogeneity is resolved upon differentiation

Dynamic and heterogeneous DNA methylation in pluripotent cells; Steffen Rulands; University of Cambridge, UK

  • Auto catalytic methylation and time delayed de methylation predicts autonomous oscillations in methylation patterns
  • Simple simulations recapitulate observations
  • Prove model correct ie oscillations true by synchronizing cells and measuring methylation over time
  • See oscillations of methylation most strikingly in enhancers but also in other elements
  • Competitive binding explains observed cpg density dependence

High-throughput, spatially resolved, single-cell transcriptomics with MERFISH; Jeffrey Moffitt; Harvard University, USA

  • Papers: Moffitt PNAS 2016, Chen science 2015
  • Multiples by binary barcodes
  • 16 rounds of imaging for whole tanscriptome in theory but often read 1s as 0s due to dropout
  • Use error robust barcodes, need at least x errors to be mistaken for another mRNA
  • In a day do single optical slice about one millimeter thick
  • Previously requires photo bleaching which is time costly
  • Now use disulfide linker that can just be washed away
  • No photo leaching also means imagining larger areas
  • Multicolor imaging now like seqFISH (only 2 colors)
  • Now 40000 cells in 18 hours
  • 75 million rna across 110000 cells overall
  • Look at spatial correlation in culture
  • Moving towards whole tissue: hypothalamus
  • There are rnas that can’t be targeted since too small, currently can capture 50% of transcriptome species

Massively parallel clonal analysis using CRISPR/Cas9 induced genetic scars; Jan Philipp Junker; Max Delbrueck Center for Molecular Medicine, Germany

  • Approaches for lineage tracking: brainbow or live tracking but it companionable with sequencing
  • Use scar sequences induced by crispr into gfp transgender to track lineage
  • Certain scars more likely than others since preferred by Dna repair mechanism
  • Cut off zebra fish tail and regenerate; lineage tracing suggest regrowth composed of same cells in same spatial organization

Dissecting cell fate choice using single-cell genomics; John Marioni; EMBL-EBI/WTSI, UK

  • BASICS plos comp model 2015 for measuring over dispersion of genes integrating spike in
  • Genome biology 2016 paper to characterize differential variability
  • Basics now adapted to be without spike ins
  • Apply to aging immune system as unstimulated T cells
  • Find younger mice have greater mean expression of immune activity and mor variability in older mice
  • Deconvolution approach by pooling single cells(?) For better diff exp(?)

MAGIC: A Diffusion based data imputation method reveals progressions and gene-gene interactions in breast cancer cells undergoing EMT; David van Dijk; Columbia University, UK

  • Recover data by learning and exchanging information from similar cells
  • Challenging to find biologically similar cells
  • Want to find manifold distance instead of Euclidean distance
  • Learn global structure through diffusion or small random walks
  • Convert from cell cell distance matrix to Markova affinity matrix
  • Diffuse data from cell to cell so that each cell has information from neighbor
  • Validate by artificially creating dropouts
  • Unclear if diffusion approach artificially connects distinct clusters?
  • Robustness to missing paths or groups in the manifold?

Single-cell spatial reconstruction reveals global division of labor in the mammalian liver; Rom Shenhav; Weizmann Institute of Science, Israel

  • Key liver functions ar spatially zoned
  • Need high resolution spatial reconstruction of zonation patterns -> smFish
  • Manually annotate central vein and draw circles to assess gradient of expression as function of distance from center
  • Quantify landmark highly expressed genes with distinct zonation profiles with smFish
  • Use single cell rnaseq for other genes
  • Infer lobule layer using expression of landmark genes

Accurate identification of somatic stem cells using single-cell RNA-sequencing; Petra Schwalie; EPFL, Switzerland

  • How many kinds of adult stem cells are there lots of remaining questions
  • Find stemness signature
  • Machine learning approach using raining on neuronal and hematopoietic stem cells , reg logistic regression
  • Apply to independent intestinal stm cell data from grun et al nature 2015
  • Use stem checker to find genes identified have previously implicated in stemness programs