PAMR-SC Calogero Raffaele & Alessandr Luca University of Torino - - PowerPoint PPT Presentation

PAMR-SC

Calogero Raffaele & AlessandrÌ Luca

University of Torino (Italy) raffaele.calogero@unito.it

SubpopulaFon OrganizaFon Of Innate T-cells

• We have a collaboraFon with Prof. De Libero, Basel

University, for the analysis of single cell data to detect the presence of sub-populaFons within 4 cell populaFons, ~500 cells/group:

– AcFvated T cells expressing the TCR γδ;

– ResFng TCR γδ cells; – AcFvated MAIT cells, – ResFng MAIT cells.

• Aim:

– IdenFfy a set of gene products, that can be use in mulF- color FACS analysis to study the detected sub-populaFons.

Prepare cells Sort cells Lysis, RT, AmplificaFon Library preparaFon Sequence Pooled single cell cDNA library HiSeq 2000 Single cell cDNA

Smart-seq2* & Single Cell FACS SorFng, 96 cells/plate

BD FACSAria II™ Special Order System

*Adapted from Picelli et al. Nature Methods 10, 1096–1098 (2013)

Frequency 2 4 6 8 10 100 200 300 400

Workflow

• Data reducFon.
• Cluster data to idenFfy sub-populaFons
• Checking robustness of the sub-populaFon

clusters

• SelecFng the main gene players in clusters

formaFon.

Data reducFon

QC pass QC fail Simpson diversity 200 400 600 800 QC pass QC fail Preseq coverage

• 1
• 0.5

0.5 1

Lorenz-staFsFc Good-Turing model of sample coverage

*Diaz et al. BioinformaFcs (2016) 32 (14): 2219-2220

#Daley et al. BioinformaFcs (2014) 30 (22): 3159-65

Data reducFon

DetecFng cell heterogeneity

Bo Wanget al Biorxiv.org doi: hlp://dx.doi.org/10.1101/052225 hlp://bioconductor.org/packages/release/bioc/html/SIMLR.html

DetecFng cell heterogeneity

Bo Wanget al Biorxiv.org doi: hlp://dx.doi.org/10.1101/052225 hlp://bioconductor.org/packages/release/bioc/html/SIMLR.html

CriFcal points in data analysis

• DetecFng cell heterogeneity:

– Defining the opFmal number of clusters. – Stability of the clusters.

• DetecFng the main players, i.e. genes, in

cluster organizaFon:

– Stability of the clusters.

Clusters stability: cells effect

−40 −20 20 −10 −5 5 10 15 20 SIMLR component 1 SIMLR component 2 MA90 MA19 MA238 MA39

1 3 2 4 5 6 bootstrap = 10% of the cells

0-25% 25-50% 50-75% 75-100%

• Allows the evaluaFon of cluster stability and

selecFon of the opFmal number of clusters.

• Scoring cluster quality:

– normalized mutual informaFon (NMI)

• Bootstrap procedure was implemented using

BiocParallel.

Cells bootstrap

DetecFng the main gene players

Feature selecFon

We use the Index of dispersion D for each gene i in cluster j (1) and in all data set (2) : We increase of an amount Si , % of Dij, the Dij and if ΔDij ≤ 0 the gene is purged: We run SIMLR on the purged genes set and we esFmate the cluster stability with respect to full dataset.

C1 C2 C3 C5

• 5

5

HCL of 194 genes Mait acFvated

C6

On-going

• Finalizing package …..

Francesca Zolezzi (Galderma) Josephine Lum Bhairav Paleja Gennaro De Libero Lucia Mori Raffaele Calogero Luca AlessandrÌ