当前位置:网站首页>Topic 1 Single_Cell_analysis(2)
Topic 1 Single_Cell_analysis(2)
2022-06-12 07:38:00 【柚子味的羊】
Topic 1 Single_Cell_analysis(2)
文章目录
Part 1 Scissor——logistic regression
1.下载bulk-dataset和phenotype(TP53mutant作为phenotype)
load(url(paste0(location, 'TCGA_LUAD_exp2.RData')))
load(url(paste0(location, 'TCGA_LUAD_TP53_mutation.RData')))
table(TP53_mutation)

共有497个cells,其中255个是mutant,剩下的243个是wild-type(mutant设置为1,wild-type设置为0)
2.Scissor选择有信息量的细胞(informative cells)
#使用scissor选择信号细胞(TP53 mutant==1 wild-ttype==0)family=binormal
phenotype<-TP53_mutation
tag <- c("wild-type", "TP53 mutant")
infos3<-Scissor(bulk_dataset,sc_dataset,phenotype,tag=tag,alpha=NULL,family = "binomial",Save_file="Scissor_LUAD_TP53_mutation.RData")
此处不设置其他参数值,最终自定义!
由上图可知,当alpha=0.2时,识别到414个mutant cell和320个wild-type cell,将识别出的cell与全部细胞呈现在picture
Scissor_select<-rep(0,ncol(sc_dataset))
names(Scissor_select)<-colnames(sc_dataset)
Scissor_select[infos3$Scissor_pos]<- 1
Scissor_select[infos3$Scissor_neg]<- 2
sc_dataset<-AddMetaData(sc_dataset,metadata = Scissor_select,col.name = "scissor")
DimPlot(sc_dataset,reduction = 'umap',group.by = 'scissor',cols = c('grey','indianred1','royalblue'),pt.size = 1.2,order = c(2,1))

Sissor_pos:表示与mutant紧密相关的cell(因为在设置的过程中设置mutant为1)
Scissor_neg:表示wild-type紧密相关的cell(在设置过程中设置wild-type为0)
Part 2 Scissor 结果解释
Scissor_postive 细胞和 Scissor_negative 细胞都是与特定表型(如TP53mutantion、survival等)高度相关的 Scissor 选择细胞。Scissor选择的细胞和表型之间的关联取决于使用的模型,应该在特定的上下文中进行解释。对于线性回归和分类模型,参数的初始值phenotype都会影响解释。
例如,在上述使用TP53突变状态的应用中,如果TP53突变体编码为1,而野生型为0,则 Scissor+ 细胞将与TP53突变体相关联,而 Scissor- 细胞将与野生型相关联。如果编码颠倒phenotype对于两种表型,Scissor+细胞和Scissor-细胞的解释相应改变。对于 Cox 回归,Scissor+ 细胞与较差的存活率相关,而 Scissor- 细胞与良好的存活率相关。
Scissor可以将细胞与表型联系起来,这种联系是表型之间的相对概念。也就是说,**Scissor 分配了一个细胞比另一个表型更可能与哪个表型相关联。**考虑到单细胞和体样本之间可能存在负相关,我们可以通过将一个细胞分为以下三类(表 1)来进一步解释一个细胞:**如果一个细胞与所有体样本的相关性平均值大于 0,并且正相关数大于负相关数,该细胞与相关表型更相似;如果一个细胞相关性的平均值小于 0 并且负相关性的数量大于正相关性的数量,则该细胞应被解释为与其他表型更不相似;否则,该细胞与表型的关联是无法确定的。**在大多数情况下,负相关值非常少,识别出的细胞属于“更相似”的类别。
注:Development
1.label
在所有的数据中经过Scissor只能将数据处理成Scissor_pos和Scissor_neg两个label——>多标签
2.model of selected cell and phenotype
Scissor选择出的positive和negative细胞,根据自己建立的model在cell和phenotype的关联中选择的(特定上下文选择)查:关联度选择
Question
1. the relationship of single-cell & sample
单细胞和体样本之间可能存在负相关 ??
边栏推荐
- The function of C language string Terminator
- Arrangement of statistical learning knowledge points gradient descent, least square method, Newton method
- Model deployment learning notes (I)
- Voice assistant - potential skills and uncalled call technique mining
- Continuous local training for better initialization of Federated models
- LeetCode34. 在排序数组中查找元素的第一个和最后一个位置
- Unity uses shaders to highlight the edges of ugu I pictures
- Velocity autocorrelation function lammps v.s MATALB
- Gd32f4 (5): gd32f450 clock is configured as 200m process analysis
- Embedded gd32 code read protection
猜你喜欢

Embedded gd32 code read protection

Paddepaddl 28 supports the implementation of GHM loss, a gradient balancing mechanism for arbitrary dimensional data (supports ignore\u index, class\u weight, back propagation training, and multi clas

最新hbuilderX编辑uni-app项目运行于夜神模拟器

Logistic regression

Keil installation of C language development tool for 51 single chip microcomputer

GD32F4(5):GD32F450时钟配置为200M过程分析

5 lines of code identify various verification codes

Golang quickly generates model and queryset of database tables

SQL -- course experiment examination

Set up a remote Jupiter notebook
随机推荐
Chapter 4 - key management and distribution
MySQL index (easy to handle in one article)
Unity uses shaders to highlight the edges of ugu I pictures
Arrangement of statistical learning knowledge points -- maximum likelihood estimation (MLE) and maximum a posteriori probability (map)
Generalized semantic recognition based on semantic similarity
Qt实现托盘
Right click the general solution of file rotation jam, refresh, white screen, flash back and desktop crash
R语言使用caTools包的sample.split函数将机器学习数据集划分为训练集和测试集
Summary of software testing tools in 2021 - unit testing tools
Golang 快速生成数据库表的 model 和 queryset
[college entrance examination] prospective college students look at it, choose the direction and future, and grasp it by themselves
ECMAScript6面试题
Summary of machine learning + pattern recognition learning (IV) -- decision tree
R语言使用neuralnet包构建神经网络回归模型(前馈神经网络回归模型),计算模型在测试集上的MSE值(均方误差)
VS2019 MFC IP Address Control 控件继承CIPAddressCtrl类重绘
What is a good recommendation system?
QT realization tray
R语言dplyr包mutate_at函数和one_of函数将dataframe数据中指定数据列(通过向量指定)的数据类型转化为因子类型
Missing getting in online continuous learning with neuron calibration thesis analysis + code reading
Study on display principle of seven segment digital tube