In githubz0r/SecretUtils: Some utils for Conos

load packages etc.

library(conos)
require(pagoda2)
devtools::load_all('/home/larsc/SecretUtils')
library(tidyverse)
library(cowplot)

Epilepsy

epilepsy_con <- readRDS(file.path('/home/larsc/data/10x_preproced_graphed.rds'))
epilepsy_annot <- readRDS(file.path('/home/demharters/R/projects/UPF9_14_17_19_22_23_24_32_33/metadata_10x_final.rds'))

Fix annotation a little bit

epilepsy_annot$cellid <- rownames(epilepsy_annot)

nr of cells in each subtype for each condition

condition_split <- split(epilepsy_annot$subtype, epilepsy_annot$condition, drop=TRUE)
condition_split %>% lapply(table)

Read unaligned graph adj obtained with pagoda2 on the whole raw matrix

eps_unaligned_adj <- readMM('/home/larsc/data/eps_unaligned_adj.mtx')

paga_subtype_condition <- GeneratePagaItems(eps_unaligned_adj, epilepsy_annot$subtype, epilepsy_annot$condition,
                                                     by.subtypes.condition = T, log.scale=F)
paga_subtype_condition$scatter.plot # has matrix, data frame and plot

Distance matrix tsne plot, using only a single n_sample x n_sample matrix, using log scale

samples_connectivities <- GeneratePagaItems(eps_unaligned_adj, sample.vector=epilepsy_annot$sample, by.sample=T, log.scale=T, 
                                            pseudo.connectivity = 1e-3)
# put it in log scale for better visualization, so we add a small pseudo connectivity
SecretUtils::PlotDistanceMatRed(samples_connectivities$connectivities, epilepsy_annot$sample, epilepsy_annot$subtype,
                                epilepsy_annot$sample, epilepsy_annot$cellid, epilepsy_annot$condition, perplexity=2, 
                                max_iter=1e5, by.subtype=F)

Within conditions distances

samples_connectivities <- GeneratePagaItems(eps_unaligned_adj, sample.vector=epilepsy_annot$sample, by.sample=T)
SecretUtils::ConditionDistanceDensity(samples_connectivities$connectivities, epilepsy_annot$sample, epilepsy_annot$subtype, 
                                      epilepsy_annot$sample, epilepsy_annot$cellid, epilepsy_annot$condition, notch=F, by.subtype=F)

Using subtype-sample partitions

paga_subtype_samples <- GeneratePagaItems(eps_unaligned_adj, epilepsy_annot$subtype, epilepsy_annot$condition, 
                                          epilepsy_annot$sample, by.subtypes.samples = T)

paga_subtype_samples$sub.cond.plot

Deeper look at between distances

GeneratePagaPlot(paga_subtype_samples$paga.df, subset='between')

log scale

paga_subtype_samples_unlin <- GeneratePagaItems(eps_unaligned_adj, epilepsy_annot$subtype, epilepsy_annot$condition, 
                                          epilepsy_annot$sample, by.subtypes.samples = T, log.scale=T, 
                                          pseudo.connectivity=1e-3)

paga_subtype_samples_unlin$sub.cond.plot
GeneratePagaPlot(paga_subtype_samples_unlin$paga.df, subset='between', log.scale=T)

Pancreas indrop, 4 samples

pancreas_con <- readRDS('/home/larsc/data/pancreas_indrop_conos_precced_graphed.rds')
location_of_annot <- '/d0-mendel/home/viktor_petukhov/SmallProjects/conos_back/data/conditions/seurat_islets/indrop/design_info.csv'
pancreas_annot <- location_of_annot %>% data.table::fread(sep=",",header=T) %>% as.data.frame()

nr of cells in each subtype for each condition, lots of very small partitions so they will be distorting the results

condition_split <- split(pancreas_annot$CellType, pancreas_annot$Disease, drop=TRUE)
condition_split %>% lapply(table)

Create unaligned matrix

raw_pancreas <- RbindRaw(pancreas_con)
unaligned_pancreas <- SecretUtils::GenerateUnalignedAdj(raw_pancreas, pancreas_annot$CellId)

Scatter plot of PAGA values for subtype-condition partitions

paga_subtype_condition <- GeneratePagaItems(unaligned_pancreas, pancreas_annot$CellType, pancreas_annot$Disease,
                                                     by.subtypes.condition = T)
paga_subtype_condition$scatter.plot

Distance matrix tsne plot, using only a single n_sample x n_sample matrix.

samples_connectivities <- GeneratePagaItems(unaligned_pancreas, sample.vector=pancreas_annot$Individual, by.sample=T)
# put it in log scale for better visualization, so we add a small pseudo connectivity
SecretUtils::PlotDistanceMatRed(log(samples_connectivities$connectivities+1e-6), pancreas_annot$Individual, pancreas_annot$CellType,
                                pancreas_annot$Individual, pancreas_annot$CellId, pancreas_annot$Disease, perplexity=1, 
                                max_iter=1e5, by.subtype=F)

Within conditions distances using same single sample x sample matrix (we only have 1 diseased sample/individual)

SecretUtils::ConditionDistanceDensity(samples_connectivities$connectivities, pancreas_annot$Individual, pancreas_annot$CellType, 
                                      pancreas_annot$Individual, pancreas_annot$CellId, pancreas_annot$Disease, notch=F,
                                      by.subtype=F)

paga_subtype_samples <- GeneratePagaItems(unaligned_pancreas, pancreas_annot$CellType, pancreas_annot$Disease, 
                                          pancreas_annot$Individual, by.subtypes.samples = T)

paga_subtype_samples$sub.cond.plot

closer look at between distances

GeneratePagaPlot(paga_subtype_samples$paga.df, subset='between')

githubz0r/SecretUtils documentation built on May 15, 2021, 10:29 p.m.