主成分分析(PCA)是一种无监督降维方法,能够有效对高维数据进行处理。但PCA对相关性较小的变量不敏感,而PLS-DA(偏最小二乘判别分析)能够有效解决这个问题。而OPLS-DA(正交偏最小二乘判别分析)结合了正交信号和PLS-DA来筛选差异变量。
“本分析主要用于代谢组学中差异代谢物的筛选。
液相色谱高分辨质谱法(LTQ Orbitrap)分析了来自183位成人的尿液样品。
sacurine
list 包含了三个数据矩阵:
dataMatrix
为样本-代谢物含量矩阵(log10转换过),记录了各种类型的代谢物在各样本中的含量信息。共计183个样本(行)以及109种代谢物(列)。
sampleMetadata
中记录了183个样本所来源个体的年零、体重、性别等信息。
variableMetadata
为109种代谢物的注释详情,MSI level水平。
rm(list = ls())
# load packages
library(ropls)
# load data
data(sacurine)
#查看数据集
head(sacurine$dataMatrix[ ,1:2])
head(sacurine$sampleMetadata)
head(sacurine$variableMetadata)
#提取性别分类
genderFc = sampleMetadata[, "gender"]
> head(sacurine$dataMatrix[ ,1:2])
(2-methoxyethoxy)propanoic acid isomer (gamma)Glu-Leu/Ile
HU_011 3.019766 3.888479
HU_014 3.814339 4.277149
HU_015 3.519691 4.195649
HU_017 2.562183 4.323760
HU_018 3.781922 4.629329
HU_019 4.161074 4.412266
> head(sacurine$sampleMetadata)
age bmi gender
HU_011 29 19.75 M
HU_014 59 22.64 F
HU_015 42 22.72 M
HU_017 41 23.03 M
HU_018 34 20.96 M
HU_019 35 23.41 M
# 分组以性别为例
# 通过orthoI指定正交组分数目
# orthoI = NA时,执行OPLS,并通过交叉验证自动计算适合的正交组分数
oplsda = opls(dataMatrix, genderFc, predI = 1, orthoI = NA)
OPLS-DA
183 samples x 109 variables and 1 response
standard scaling of predictors and response(s)
R2X(cum) R2Y(cum) Q2(cum) RMSEE pre ort pR2Y pQ2
Total 0.275 0.73 0.602 0.262 1 2 0.05 0.05
Snipaste_2021-10-28_21-32-57
结果中,R2X
和R2Y
分别表示所建模型对X和Y矩阵的解释率,Q2
表示模型的预测能力,它们的值越接近于1表明模型的拟合度越好,训练集的样本越能够被准确划分到其原始归属中。
Inertia(惯量)柱形图(左上)
展示了3个正交轴的R2Y
和Q2Y
。通过展示累计解释率评估正交组分是否足够。
显著性诊断(右上)
实际和模拟模型的R2Y
和Q2Y
值经随机排列后的散点图,模型R2Y
和Q2Y
(散点)大于真实值时(横线),表明产生过拟合2。右上图,OPLS-DA模型的R2Y和Q2Y与随机置换数据后获得的相应值进行比较。
离群点展示(左下)
展示了各样本在投影平面内以及正交投影面的距离,具有高值的样本标注出名称,表明它们与其它样本间的差异较大。颜色代表性别分组。
x-score plot(右下)
各样本在OPLS-DA轴中的坐标,颜色代表性别分组。
library(ggplot2)
library(ggsci)
library(tidyverse)
#提取样本在 OPLS-DA 轴上的位置
sample.score = oplsda@scoreMN %>% #得分矩阵
as.data.frame() %>%
mutate(gender = sacurine[["sampleMetadata"]][["gender"]],
o1 = oplsda@orthoScoreMN[,1]) #正交矩阵
head(sample.score)#查看
> head(sample.score)
p1 gender o1
HU_011 -1.582933 M -4.9806037
HU_014 1.372806 F -1.7443382
HU_015 -3.341370 M -3.4372771
HU_017 -3.590063 M -0.9794960
HU_018 -1.662716 M 0.3155845
HU_019 -2.312923 M 0.6561281
p <- ggplot(sample.score, aes(p1, o1, color = gender)) +
geom_hline(yintercept = 0, linetype = 'dashed', size = 0.5) + #横向虚线
geom_vline(xintercept = 0, linetype = 'dashed', size = 0.5) +
geom_point() +
#geom_point(aes(-10,-10), color = 'white') +
labs(x = 'P1(5.0%)',y = 'to1') +
stat_ellipse(level = 0.95, linetype = 'solid',
size = 1, show.legend = FALSE) + #添加置信区间
scale_color_manual(values = c('#008000','#FFA74F')) +
theme_bw() +
theme(legend.position = c(0.1,0.85),
legend.title = element_blank(),
legend.text = element_text(color = 'black',size = 12, family = 'Arial', face = 'plain'),
panel.background = element_blank(),
panel.grid = element_blank(),
axis.text = element_text(color = 'black',size = 15, family = 'Arial', face = 'plain'),
axis.title = element_text(color = 'black',size = 15, family = 'Arial', face = 'plain'),
axis.ticks = element_line(color = 'black'))
p
Snipaste_2021-10-28_22-49-44
#VIP 值帮助寻找重要的代谢物
vip <- getVipVn(oplsda)
vip_select <- vip[vip > 1] #通常以VIP值>1作为筛选标准
head(vip_select)
vip_select <- cbind(sacurine$variableMetadata[names(vip_select), ], vip_select)
names(vip_select)[4] <- 'VIP'
vip_select <- vip_select[order(vip_select$VIP, decreasing = TRUE), ]
head(vip_select) #带注释的代谢物,VIP>1 筛选后,并按 VIP 降序排序
> head(vip_select)
msiLevel hmdb chemicalClass
p-Anisic acid 1 HMDB01101 AroHoM
Malic acid 1 HMDB00156 Organi
Testosterone glucuronide 2 HMDB03193 Lipids:Steroi
Pantothenic acid 1 HMDB00210 AliAcy
Acetylphenylalanine 1 HMDB00512 AA-pep
alpha-N-Phenylacetyl-glutamine 1 HMDB06344 AA-pep
VIP
p-Anisic acid 2.533220
Malic acid 2.479289
Testosterone glucuronide 2.421591
Pantothenic acid 2.165296
Acetylphenylalanine 1.988311
alpha-N-Phenylacetyl-glutamine 1.965807
#对差异代谢物进行棒棒糖图可视化
#代谢物名字太长进行转换
vip_select$cat = paste('A',1:nrow(vip_select), sep = '')
p2 <- ggplot(vip_select, aes(cat, VIP)) +
geom_segment(aes(x = cat, xend = cat,
y = 0, yend = VIP)) +
geom_point(shape = 21, size = 5, color = '#008000' ,fill = '#008000') +
geom_point(aes(1,2.5), color = 'white') +
geom_hline(yintercept = 1, linetype = 'dashed') +
scale_y_continuous(expand = c(0,0)) +
labs(x = '', y = 'VIP value') +
theme_bw() +
theme(legend.position = 'none',
legend.text = element_text(color = 'black',size = 12, family = 'Arial', face = 'plain'),
panel.background = element_blank(),
panel.grid = element_blank(),
axis.text = element_text(color = 'black',size = 15, family = 'Arial', face = 'plain'),
axis.text.x = element_text(angle = 90),
axis.title = element_text(color = 'black',size = 15, family = 'Arial', face = 'plain'),
axis.ticks = element_line(color = 'black'),
axis.ticks.x = element_blank())
p2
Snipaste_2021-10-28_23-35-09
OPLS-DA在R语言中的实现 | 小蓝哥的知识荒原 (blog4xiang.world)
R包ropls的偏最小二乘判别分析(PLS-DA)和正交偏最小二乘判别分析(OPLS-DA)
用PLS和OPLS分析代谢组数据 - 简书 (jianshu.com)
ropls: PCA, PLS(-DA) and OPLS(-DA) for multivariate analysis and feature selection of omics data (bioconductor.org)
单组学的多变量分析|1.PCA和PLS-DA
单组学的多变量分析| 2.稀疏偏最小二乘判别分析(sPLS-DA)