一个长大了才会遇到的帅哥，
稳健，潇洒，大方，靠谱。
一段生信缘，一棵技能树。
生信技能树核心成员，单细胞天地特约撰稿人，简书创作者，单细胞数据科学家。
Identification of spatial expression trends in single-cell gene expression data
空间转录组技术使得我们可以在组织成像的基础上考察基因表达情况，同时也需要新的分析策略。trendsceek是一种基于标记点过程的方法，识别具有显著空间表达趋势的基因。trendsceek在空间转录组和顺序荧光原位杂交数据中都能很好地发现空间差异基因，并在单细胞RNA-seq数据的低维投影（TSNE/umap）中揭示了显著的基因表达梯度和热点。
代码语言：javascript复制library(trendsceek)
library(Seurat)
library(SeuratData)
AvailableData()
stxBrain.SeuratData::anterior1 -> sto
head(sto@images$anterior1@coordinates)
tissue row col imagerow imagecol
AAACAAGTATCTCCCA-1 1 50 102 7475 8501
AAACACCAATAACTGC-1 1 59 19 8553 2788
AAACAGAGCGACTCCT-1 1 14 94 3164 7950
AAACAGCTTTCAGAAG-1 1 43 9 6637 2099
AAACAGGGTCTATATT-1 1 47 13 7116 2375
AAACATGGTGAGAGGA-1 1 62 0 8913 1480
pp = pos2pp(sto@images$anterior1@coordinates[,c(2,3)])
log.fcn = log10
counts_sub[1:2,1:4]
pp = set_marks(pp, as.matrix(sto@assays$Spatial@counts), log.fcn = log.fcn)
min.ncells.expr = 3
min.expr = 5
counts_filt = genefilter_exprmat(as.matrix(sto@assays$Spatial@counts), min.expr, min.ncells.expr)
dim(counts_filt)
quantile.cutoff = 0.9 ##filter out the most lowly expressed genes from the fitting
method = 'glm' ##For (robust) linear regression set to 'rlm'
vargenes_stats = calc_varstats(counts_filt, counts_filt, quant.cutoff = quantile.cutoff, method = method)
n.top2plot = 10
topvar.genes = rownames(vargenes_stats[['real.stats']])[1:n.top2plot]
pp2plot = pp_select(pp, topvar.genes)
plot.ercc.points = FALSE
plot_cv2vsmean(vargenes_stats, topvar.genes, plot.ercc.points = plot.ercc.points)
min.count = 1
counts_norm = deseq_norm(as.matrix(sto@assays$Spatial@counts), min.count)
counts_sub = counts_norm[topvar.genes, ]
dim(counts_sub)
plot_pp_scatter(pp2plot, log_marks = FALSE, scale_marks = FALSE, pal.direction = -1)
nrand = 100
ncores = 1
##run
trendstat_list = trendsceek_test(pp2plot, nrand, ncores)
trendstat_list
head(trendstat_list$sig_genes_list$Vmark)
gene test earlystop max.env.rel.dev max.rel.dev min.pval nsim_max nsim_stop p.bh p.bo rank
S100a5 S100a5 Vmark 0 6.898791 0.29728032 0.00990099 2 2 0.0110011 0.0990099 1
Fabp7 Fabp7 Vmark 0 5.392828 0.12836321 0.00990099 2 2 0.0110011 0.0990099 2
Ptgds Ptgds Vmark 0 3.491384 0.09823452 0.00990099 2 2 0.0110011 0.0990099 3
Clca3a1 Clca3a1 Vmark 0 3.075842 0.35753230 0.00990099 2 2 0.0110011 0.0990099 4
Ttr Ttr Vmark 0 2.962141 0.10187457 0.00990099 2 2 0.0110011 0.0990099 5
Kl Kl Vmark 0 1.762761 0.11802672 0.00990099 2 2 0.0110011 0.0990099 6
alpha = 0.05 ##Benjamini-Hochberg
sig_list = extract_sig_genes(trendstat_list, alpha)
lapply(sig_list, nrow)
sig_genes = sig_list[['markcorr']][, 'gene']
plot_trendstats(trendstat_list, sig_genes[1])
plot_pp_scatter(pp_sig, log_marks = FALSE, scale_marks = FALSE, pal.direction = -1,pointsize.factor = 1)
References
[1] https://github.com/edsgard/trendsceek
[2] Edsgärd D. et al., Identification of spatial expression trends in single-cell gene expression data, Nature Methods, 2018: doi:10.1038/nmeth.4634
