Part_C_template - Stella Dushime
####Run through this example and try to understand what is going on with the data
####So, lets load RMarkdown
library(rmarkdown)set messages to FALSE on everything (prevents certain boring things from being shown in the results)
knitr::opts_chunk$set(echo = TRUE, message=FALSE,warning=FALSE,collapse = TRUE)PACKAGES
library(reshape2)
library(ggplot2)
library(dplyr)
library(plotly)
library(viridis)
library(data.table)
library(pheatmap)
library(tidyverse)
library(ggthemes)
library(clipr)
library(tidyr)
library(Rcpp)COLORS
mycolors<-c(viridis(15))
felix_cols<-mycolors[c(5,2)]
felix_4cols<-mycolors[c(15,10,8,2)]
plain_cols1<-c("blue","gray")
plain_cols2<-c("red","gray")
pats_cols<-colorRampPalette(c("#FDE725FF", "white","#440154FF"))(21)
leos_cols<-colorRampPalette(c("white","blue"))(10)
LOAD DATA AND MAKE AN OVERALL HEATMAP
## load the dataset
BCC<-read.csv("breast_cancer_cells.csv")
#and then view it
view(BCC)now let’s visualize the dataset and look for initial trends. We can do this by making a matrix so and then a heatmap to visualize the data
#make a matrix of just raw the data
BCC_MAT<-BCC%>% select(MCF10A_1:SKBR3_2)%>% round(.,02)
# what does that look like?
head(BCC_MAT)
## MCF10A_1 MCF10A_2 MCF7_1 MCF7_2 MDA231_1 MDA231_2 MDA468_1 MDA468_2 SKBR3_1
## 1 9.54 4.58 5.07 5.42 25.43 27.42 4.56 3.88 8.46
## 2 14.00 11.58 6.49 6.64 9.80 10.31 6.84 8.75 11.91
## 3 10.22 8.29 11.55 10.82 12.48 10.11 9.54 8.46 9.10
## 4 9.00 6.35 13.40 14.82 14.94 11.33 6.87 7.92 8.54
## 5 8.21 4.44 12.08 9.82 16.51 12.34 15.43 8.50 7.93
## 6 12.51 15.84 8.05 8.38 6.78 8.46 4.48 7.18 13.27
## SKBR3_2
## 1 5.64
## 2 13.67
## 3 9.43
## 4 6.82
## 5 4.75
## 6 15.05## make a heatmap from the data
pheatmap(BCC_MAT, color=pats_cols,cellwidth=30,cellheight=.03,cluster_cols=FALSE,cluster_rows=TRUE,legend=TRUE,fontsize = 7,scale="column")
##INTERPRETATION##
## What can you see in this figure?(gene expression levels across different breast cancer cell lines) are the repeated measures/reps similar or different?(the repeated measures are similar) What does this say about the precision and accuracy of them? (accuracy cant be determined but similarity indicates high precision)
##How does the control compare to the variables? (the control line is distinct (MCF10A)) Is this what you might expect? Why?(yes,because normal and cancer cells have different gene expressions) What would you look for in the literature to support this idea? (research papers and studies showing the origin and details of these cell lines)
#looks like MCF10A line is different than the other cells lines. We will take a closer look at that in the next section, once we've done some further data manipulations
Data Manipulation
## first, make new columns that combine the two reps for each variable by averaging them
BCC.2<- BCC %>% mutate(mean_MCF10A= ((MCF10A_1+MCF10A_2)/2),mean_MCF7= ((MCF7_1+MCF7_2)/2),mean_MDA231= ((MDA231_1+MDA231_2)/2),mean_MDA468= ((MDA468_1+MDA468_2)/2),mean_SKBR3= ((SKBR3_1+SKBR3_2)/2))
#did it create your new columns? YES
head(BCC.2)
## Gene_Symbol
## 1 NES
## 2 ZC3HC1
## 3 CAMSAP1L1
## 4 COBRA1
## 5 HAUS6
## 6 CHCHD4
## Description
## 1 Nestin OS=Homo sapiens GN=NES PE=1 SV=2
## 2 Isoform 2 of Nuclear-interacting partner of ALK OS=Homo sapiens GN=ZC3HC1
## 3 Isoform 2 of Calmodulin-regulated spectrin-associated protein 2 OS=Homo sapiens GN=CAMSAP1L1
## 4 Negative elongation factor B OS=Homo sapiens GN=COBRA1 PE=1 SV=1
## 5 Isoform 3 of HAUS augmin-like complex subunit 6 OS=Homo sapiens GN=HAUS6
## 6 Isoform 2 of Mitochondrial intermembrane space import and assembly protein 40 OS=Homo sapiens GN=CHCHD4
## Peptides MCF10A_1 MCF10A_2 MCF7_1 MCF7_2 MDA231_1 MDA231_2 MDA468_1 MDA468_2
## 1 7 9.54 4.58 5.07 5.42 25.43 27.42 4.56 3.88
## 2 2 14.00 11.58 6.49 6.64 9.80 10.31 6.84 8.75
## 3 5 10.22 8.29 11.55 10.82 12.48 10.11 9.54 8.46
## 4 3 9.00 6.35 13.40 14.82 14.94 11.33 6.87 7.92
## 5 5 8.21 4.44 12.08 9.82 16.51 12.34 15.43 8.50
## 6 5 12.51 15.84 8.05 8.38 6.78 8.46 4.48 7.18
## SKBR3_1 SKBR3_2 pvalue_MCF7_vs_MCF10A pvalue_MDA231_vs_MCF10A
## 1 8.46 5.64 0.54152658 0.01849512
## 2 11.91 13.67 0.03589005 0.15732968
## 3 9.10 9.43 0.20243066 0.31440425
## 4 8.54 6.82 0.05045427 0.13484982
## 5 7.93 4.75 0.16972788 0.10216410
## 6 13.27 15.05 0.07037866 0.07222737
## pvalue_MDA468_vs_MCF10A pvalue_SKBR3_vs_MCF10A mean_MCF10A mean_MCF7
## 1 0.37423781 0.9983818 7.060 5.245
## 2 0.08358401 0.9980974 12.790 6.565
## 3 0.83825377 0.9967306 9.255 11.185
## 4 0.86215724 0.9958838 7.675 14.110
## 5 0.28867772 0.9957949 6.325 10.950
## 6 0.06005626 0.9948536 14.175 8.215
## mean_MDA231 mean_MDA468 mean_SKBR3
## 1 26.425 4.220 7.050
## 2 10.055 7.795 12.790
## 3 11.295 9.000 9.265
## 4 13.135 7.395 7.680
## 5 14.425 11.965 6.340
## 6 7.620 5.830 14.160
## then make some new columns that store the log ratios of the variable means you just created, as compared to the control
BCC.2<-BCC.2 %>% mutate(log_MCF7=log2(mean_MCF7/mean_MCF10A),log_MDA231=log2(mean_MDA231/mean_MCF10A),log_MDA468=log2(mean_MDA468/mean_MCF10A),log_SKBR3=log2(mean_SKBR3/mean_MCF10A))
colnames(BCC.2)
## [1] "Gene_Symbol" "Description"
## [3] "Peptides" "MCF10A_1"
## [5] "MCF10A_2" "MCF7_1"
## [7] "MCF7_2" "MDA231_1"
## [9] "MDA231_2" "MDA468_1"
## [11] "MDA468_2" "SKBR3_1"
## [13] "SKBR3_2" "pvalue_MCF7_vs_MCF10A"
## [15] "pvalue_MDA231_vs_MCF10A" "pvalue_MDA468_vs_MCF10A"
## [17] "pvalue_SKBR3_vs_MCF10A" "mean_MCF10A"
## [19] "mean_MCF7" "mean_MDA231"
## [21] "mean_MDA468" "mean_SKBR3"
## [23] "log_MCF7" "log_MDA231"
## [25] "log_MDA468" "log_SKBR3"
#did it create your new columns? YES## ALTERNATIVE CHOICE: make a matrix of just log values the data and a heat map of that. How do the two heat maps compare? This is not possible if your values contain 0's
BCC_MAT2<-BCC.2 %>% select(log_MCF7:log_SKBR3) %>% as.matrix() %>% round(.,2)
pheatmap(BCC_MAT2, color=pats_cols,cellwidth=30,cellheight=.03,cluster_cols=FALSE,cluster_rows=TRUE,legend=TRUE,fontsize = 7,scale="column")
Diving deeping with VOLCANO PLOTS
##BASIC VOLCANO PLOT
## Add a column that stores the negative log of the pvalue of interest (in this case, SKBR3)
BCC.2<-BCC.2%>% mutate(neglog_SKBR3=-log10(pvalue_SKBR3_vs_MCF10A))
## Use ggplot to plot the log ratio of SKBR3 against the -log p-value of SKBR3
volcano_plot<-BCC.2%>%ggplot(aes(x=log_SKBR3,y=neglog_SKBR3,Description=Gene_Symbol))+geom_point(alpha=0.7,color="royalblue")
#to view it, type:
volcano_plot
## BETTER VOLCANO PLOT
BCC.2<-BCC.2%>%mutate(significance=ifelse((log_SKBR3>2&neglog_SKBR3>2.99),"UP",ifelse((log_SKBR3<c(-2)&neglog_SKBR3>2.99),"DOWN","NOT SIG")))
plain_cols3<-c("red","gray","blue")
better_volvano_plot<-BCC.2%>%ggplot(aes(x=log_SKBR3,y=neglog_SKBR3,Description=Gene_Symbol,color=significance))+geom_point(alpha=0.7)+scale_color_manual(values=plain_cols3)+xlim(-6,6)+theme_bw()+theme(axis.text = element_text(colour = "black",size = 14))+theme(text=element_text(size=14))+labs(x="log_SKBR3 compared to MCF10A",y="-log(p-value)")
better_volvano_plot
## Define the significant ones (by using ifelse and setting # parameters) so they can be colored
## Some standard colors
## volcano plot as before with some added things
#to view it, type
#check viewer and / or plots to see it
## use ggplotly to see hover over the points to see the gene names. Record these for the next step
ggplotly(better_volvano_plot)
##INTERPRETATION##
## Significant up regulated are: TDP2, KRT23 and GCLC
## Significant down regulated are:APOA1,HLA-A and MYO1B
#Why? How?Barplots of significant points of interest
EXAMPLES OF A COUPLE PROTEINS or GENES
# Make a pivot longer table of the C19 data for Control_2h_1:Virus_24h_2
BCC_long<-pivot_longer(BCC.2,cols = c(MCF10A_1:SKBR3_2), names_to ='variable')%>% select(-c(pvalue_MCF7_vs_MCF10A:significance))
head(BCC_long)
## # A tibble: 6 × 5
## Gene_Symbol Description Peptides variable value
## <chr> <chr> <int> <chr> <dbl>
## 1 NES " Nestin OS=Homo sapiens GN=NES PE=1 SV=2" 7 MCF10A_1 9.54
## 2 NES " Nestin OS=Homo sapiens GN=NES PE=1 SV=2" 7 MCF10A_2 4.58
## 3 NES " Nestin OS=Homo sapiens GN=NES PE=1 SV=2" 7 MCF7_1 5.07
## 4 NES " Nestin OS=Homo sapiens GN=NES PE=1 SV=2" 7 MCF7_2 5.42
## 5 NES " Nestin OS=Homo sapiens GN=NES PE=1 SV=2" 7 MDA231_1 25.4
## 6 NES " Nestin OS=Homo sapiens GN=NES PE=1 SV=2" 7 MDA231_2 27.4
## based on the gene symbols from plotly, select a few proteins from the table. Select just the data and gene symbols and pivot longer
Examples_Down<-BCC_long%>%filter(Gene_Symbol=="APOA1"|Gene_Symbol=="HLA-A;MYO1B"|Gene_Symbol=="HMGN5")
head(Examples_Down)
## # A tibble: 6 × 5
## Gene_Symbol Description Peptides variable value
## <chr> <chr> <int> <chr> <dbl>
## 1 HMGN5 " High mobility group nucleosome-binding … 4 MCF10A_1 24.0
## 2 HMGN5 " High mobility group nucleosome-binding … 4 MCF10A_2 24.8
## 3 HMGN5 " High mobility group nucleosome-binding … 4 MCF7_1 11.0
## 4 HMGN5 " High mobility group nucleosome-binding … 4 MCF7_2 11.0
## 5 HMGN5 " High mobility group nucleosome-binding … 4 MDA231_1 4.62
## 6 HMGN5 " High mobility group nucleosome-binding … 4 MDA231_2 5.69
## make barplots facetted by Gene Symbol (when working with other data sets - change x=order to x = variable)
Example_plot_down<-Examples_Down%>%ggplot(aes(x=factor(variable,levels = c('MCF10A_1','MCF10A_2','MCF7_1','MCF7_2','MDA231_1','MDA231_2','MDAA468_1','MDA468_2','SKBR3_1','SKBR3_2')),y= value))+geom_bar(stat = "identity",fill="red")+facet_wrap(~Gene_Symbol)+theme_bw()+theme(axis.text = element_text(colour="black",size=10))+theme(text= element_text(size=14))+theme(axis.text.x = element_text(angle=45, hjust = 1))+labs(x="sample",y="relative intensity")
Example_plot_down
#check viewer and / or plots to see it
##INTERPRETATION## ## What can you see in this figure? are the repeated measures/reps similar or different? What does this say about the precision and accuracy of them? ##How does the control compare to the variables? Is this what you might expect? Why? What would you look for in the literature to support this idea?
## Same process for the upregulated ones
Examples_up<-BCC_long %>% filter(Gene_Symbol=="FNBP1L"|Gene_Symbol=="KRT23")
head(Examples_up)
## # A tibble: 6 × 5
## Gene_Symbol Description Peptides variable value
## <chr> <chr> <int> <chr> <dbl>
## 1 KRT23 " Keratin, type I cytoskeletal 23 OS=Homo… 3 MCF10A_1 2.39
## 2 KRT23 " Keratin, type I cytoskeletal 23 OS=Homo… 3 MCF10A_2 2.49
## 3 KRT23 " Keratin, type I cytoskeletal 23 OS=Homo… 3 MCF7_1 4.03
## 4 KRT23 " Keratin, type I cytoskeletal 23 OS=Homo… 3 MCF7_2 5.53
## 5 KRT23 " Keratin, type I cytoskeletal 23 OS=Homo… 3 MDA231_1 2.89
## 6 KRT23 " Keratin, type I cytoskeletal 23 OS=Homo… 3 MDA231_2 4.76
Example_plot_up<-Examples_up %>%ggplot(aes(x=factor(variable,levels = c('MCF10A_1','MCF10A_2','MCF7_1','MCF7_2','MDA231_1','MDA231_2','MDAA468_1','MDA468_2','SKBR3_1','SKBR3_2')),y= value))+geom_bar(stat = "identity",fill="royalblue")+facet_wrap(~Gene_Symbol)+theme_bw()+theme(axis.text = element_text(colour="black",size=10))+theme(text= element_text(size=14))+theme(axis.text.x = element_text(angle=45, hjust = 1))+labs(x="sample",y="relative intensity")
Example_plot_up
#check viewer and / or plots to see it
##INTERPRETATION## ## What can you see in this figure? are the repeated measures/reps similar or different? What does this say about the precision and accuracy of them? ##How does the control compare to the variables? Is this what you might expect? Why? What would you look for in the literature to support this idea?
#interpretation HINT:insert a chunk and create two seprate lines of code that filter for your specific upregulated genes/proteins of interest and selects for only their gene symbols and descriptions. Do this for the downregulated as well. This will generate two list of the descriptors for each gene of interest, helping you understand your figures. Be sure to view it, not just ask for the head of the table generated.
WRAP UP
##now you can knit this and publish to save and share your code. Use this to work with either the brain or breast cells and the Part_C_template to complete your lab 6 ELN. ##Annotate when you have trouble and reference which line of code you need help on ## good luck and have fun!