能做拟时序分析的方法实在是太多了,这个SCORPIUS方法也是2016丢在预印本的, 文章标题是:《SCORPIUS improves trajectory inference and identifies novel modules in dendritic cell development》,有意思的是它直到现在也没有在任何期刊上面正式发表,但是都有了54个引用(截止至2022-02-02),预印本链接是:https://www./content/10.1101/079509v2
在其文章里面对流程的介绍蛮清晰的:
主要是3个步骤:
Dimensionality reduction involves calculating the correlation distance, optionally filtering out outliers, and performing multi-dimensional scaling.
Trajectory inference creates an initial path by calculating the shortest path through k cluster centres, and by iteratively fitting this path to the data using the principal curves algorithm.
During feature selection, a Random Forest is trained using the expression data to predict the ordering of cells as outputted by the principal curves
# warning: setting num_permutations to 10 requires a long time (~30min) to run! # set it to 0 and define a manual cutoff for the genes (e.g. top 200) for a much shorter execution time. gimp <- gene_importances( expression, traj$time, num_permutations = 10, num_threads = 8, ntree = 10000, ntree_perm = 1000 ) gimp$qvalue <- p.adjust(gimp$pvalue, "BH", length(gimp$pvalue)) gene_sel <- gimp$gene[gimp$qvalue < .05] expr_sel <- scale_quantile(expression[,gene_sel])
# Draw a time series heatmap time <- traj$time draw_trajectory_heatmap(expr_sel, time)
## Also show the progression groupings draw_trajectory_heatmap(expr_sel, time, progression_group=group_name)
# warning: setting num_permutations to 10 requires a long time (~30min) to run! # set it to 0 and define a manual cutoff for the genes (e.g. top 200) for a much shorter execution time. gimp <- gene_importances( expression, traj$time, num_permutations = 10, num_threads = 8, ntree = 10000, ntree_perm = 1000 ) gimp$qvalue <- p.adjust(gimp$pvalue, "BH", length(gimp$pvalue)) gene_sel <- gimp$gene[gimp$qvalue < .05] expr_sel <- scale_quantile(expression[,gene_sel])
# Draw a time series heatmap time <- traj$time draw_trajectory_heatmap(expr_sel, time)
## Also show the progression groupings draw_trajectory_heatmap(expr_sel, time, progression_group=group_name)
We thank Dr.Jianming Zeng(University of Macau), and all the members of his bioinformatics team, biotrainee, for generously sharing their experience and codes.