当只有自己发生状况时,归结给运气和自己太菜;当别人发生同样的状况时,归结给共同的客观原因,比如网速太慢、电脑太烂;当伟大的互联网告诉你有更多的人遭遇了同样的困惑,才终于投向科学,怪罪无知,无知当然是自己的无知。
状况1: GSEA-GO 同样的设定返回了不同的结果
在做 jimmy 大神布置的作业:对比 Gorilla, clusterProfiler, topGO 三种工具
作业做到 GSEA 法
## clusterProfiler GSEA GO-BP
library(clusterProfiler)
library(hgu133plus2.db)
load("genelist_ENTREZID_Decr_pVal.Rdata")
### with decreasing p.val
gseaGO_BP <- gseGO(geneList = entrezlist_dcrpv,
OrgDb = hgu133plus2.db,
keyType = "ENTREZID",
ont = "BP",
nPerm = 1000, ## 排列数
minGSSize = 5,
maxGSSize = 500,
pvalueCutoff = 0.95,
verbose = TRUE) ## 不输出结果
gseaGO_BPresult <- gseaGO_BP@result
save(gseaGO_BPresult, file = "cp_gseaGO_BPresult.Rdata")
得到的结果是下面这样的,出现了65个条目和10个条目两种,似乎还有0条,但手慢无截图
当时的想法是这样的:
可能和网络也有关系?几次输入同样的参数,得到的结果并不一样....有时是报错 no term, 迷惑desu
毕竟日常网络不好,总天真的以为5G来了一切就解决了┑( ̄Д  ̄)┍
状况2: GSEA-KEGG 同样的设定返回了不同的结果
没有错,前后呼应,这是自称资质平平送老迎新的技能树资深半席讲师遇到的状况。依然是同样的参数,不同的结果。
讨论:也许这有一定的随机性?
后来果然发现 gseKEGG()
的一个逻辑参数 seed
是可以设置的,默认值为 FALSE
, 那么就惊喜地设置为 TRUE
, 又惊喜地发现了依然不同的结果。(sad
网友们的困惑
回到最初的问题,为什么每次会不一样呢?第二个问题,seed
这个参数到底有没有用?
自己思索无果就求助于互联网(然而其实应该别瞎想直接google
大部分第一个问题基本没有已解决的下文,第二个问题倒是终于发现了这位心明眼亮的选手:
当然也有人说并没有用....
不如自己动手尝试:
set.seed(1984)
gseaGO_BP3 <- gseGO(geneList = entrezlist_FC,
OrgDb = hgu133plus2.db,
keyType = "ENTREZID",
ont = "BP",
nPerm = 1000, ## 排列数
minGSSize = 5,
maxGSSize = 500,
pvalueCutoff = 0.05,
verbose = TRUE,
seed = TRUE)
第一次:no term enriched under specific pvalueCutoff...
第二次:结果又出现辽
看来seed确实是没用的。
另一位心明眼亮的网友如是说:
再回头去看,gseGO()
和 gseKEGG()
默认的方法是 "fgsea".
We implemented GSEA algorithm proposed by Subramanian(Subramanian et al. 2005). Alexey Sergushichev implemented an algorithm for fast GSEA analysis in the fgsea(S., n.d.) package.
In DOSE(Yu et al. 2015), user can use GSEA algorithm implemented in
DOSE
orfgsea
by specifying the parameterby="DOSE"
orby="fgsea"
. By default, DOSEusefgsea
since it is much more fast.
而 fgsea 算法的文章里有几段话(反正公式是看不懂的也就看点人话猜一下这亚子
Gene set enrichment analysis is a very widely used method for analyzing gene expression data. It allows to select from an a priori defined list of gene sets those which have non-random behavior in a considered experiment.
The method has a major drawback of being relatively slow....That can be done in a straightforward manner by sampling random gene sets.
Instead of generating nm independent random gene set for each permutation and each gene set we will generate only n radom gene sets of size K.
The preranked gene set enrichment analysis takes as input two objects: an array of gene statistic values S and a list of query gene sets P. The goal of the analysis is to determine which of the gene sets from P has a non-random behavior.
Sergushichev A. An algorithm for fast preranked gene set enrichment analysis using cumulative statistic calculation[J]. BioRxiv, 2016: 060012.
大概就是说传统的GSEA方法很慢,可以通过对随机的基因集抽样解决这个问题,也就顾名思义是 a fast gene set enrichment analysis (FGSEA) 了。
与此同时四通八达的 jimmy 大神去问了神包 clusterProfiler
的作者,神通广大的Y叔——
Y叔回答:“nPerm次数多点就行 seed没用”
毕竟👇
For each p ∈ P we need to find the enrichment statistic value and to calculate a p-value of this not to be random. To calculate a p-value for gene set p we can obtain an empirical null distribution by sampling n random gene sets of the same size as p.
Sergushichev A. An algorithm for fast preranked gene set enrichment analysis using cumulative statistic calculation[J]. BioRxiv, 2016: 060012.
所以,多试试总会得到想要的。
最后,向大家隆重推荐生信技能树的一系列干货!
- 生信技能树全球公益巡讲:https://mp.weixin.qq.com/s/E9ykuIbc-2Ja9HOY0bn_6g
- B站公益74小时生信工程师教学视频合辑:https://mp.weixin.qq.com/s/IyFK7l_WBAiUgqQi8O7Hxw
- 招学徒:https://mp.weixin.qq.com/s/KgbilzXnFjbKKunuw7NVfw