R markdowns/Somalogic_deeper_metadata_analysis.Rmd

---
title: "Differential protein expression using Limma"
output:
  html_document:
    df_print: paged
---

```{r}
library(dplyr)
library(tidyr)
library(ggplot2)
library(EnhancedVolcano)

metadata <- read.csv(file ='D:/Data/Data_drive/Data/IS_Protein_data/somalogic_metadata.csv', row.names = 1)
somaEset <- readRDS('D:/Data/Data_drive/Data/IS_Protein_data/somaEset.rds')
normalised_expression <- as.data.frame(exprs(somaEset))
protein_IDs <- as.data.frame(somaEset@featureData@data)
proteins_plot <- read.csv('D:/Data/Data_drive/Data/IS_Protein_data/proteins_plot.csv', row.names = 1)

```


```{r}
### lung fibrosis
subset_metadata <- dplyr::filter(metadata, grepl("Scleroderma", category))
subset_proteins <- normalised_expression[, colnames(normalised_expression) %in% rownames(subset_metadata)]
design <- model.matrix(~ Lung_Fibrosis + 0, data = subset_metadata)
colnames(design) <- c("No","Yes")
contrastNames <- c("Yes - No")
contrasts <- makeContrasts(contrasts = contrastNames, levels=design)
limmaRes <- subset_proteins %>% lmFit(design = design)%>% contrasts.fit(contrasts) %>% eBayes()
coeffs <- coefficients(limmaRes)

stats_df <- topTable(limmaRes, number = nrow(subset_proteins))
stats_df <- merge(stats_df, protein_IDs, by = "row.names")
write.csv(stats_df, "D:/Data/Data_drive/Data/IS_Protein_data/SSc_lung_fibrosis_allproteins.csv")
writeable_SSc_lung_fibrosis <- stats_df %>% filter( P.Value <0.05) %>% filter(logFC >0.585| logFC < -0.585) %>% select(SeqId, TargetFullName,Target,UniProt,EntrezGeneSymbol,logFC,AveExpr,P.Value)
write.csv(writeable_SSc_lung_fibrosis, "D:/Data/Data_drive/Data/IS_Protein_data/SSc_lung_fibrosis.csv")

stats_df <- stats_df %>%
  arrange(P.Value) %>%
  mutate(EntrezGeneSymbol = as.character(EntrezGeneSymbol),
         label = ifelse(P.Value < 0.05 & EntrezGeneSymbol %in% head(EntrezGeneSymbol, 50), EntrezGeneSymbol, NA))

writeable_SSc_lung_fibrosis %>%
  summarise(
    logFC_greater_than_0 = sum(logFC > 0),
    logFC_less_than_0 = sum(logFC < 0))

SSc_lung_fibrosis <- EnhancedVolcano(data.frame(stats_df), x = 'logFC', y = 'P.Value',lab = stats_df$label,selectLab = stats_df$label,
    title = 'SSc Lung Fibrosis vs SSc no Lung Fibrosis proteins',
    pCutoff = 0.05,
    FCcutoff = 0.585,
    xlim = c(min(stats_df[['logFC']], na.rm = TRUE), max(stats_df[['logFC']], na.rm = TRUE)),
    ylim = c(0, max(-log10(stats_df[['P.Value']]), na.rm = TRUE)),
    pointSize = 2.0,
    labSize = 4.0,
    labCol = 'grey14',
    colAlpha = 4/5,
    boxedLabels = T,
    legendPosition = 'None',
    legendLabSize = 8,
    legendIconSize = 3.0,
    drawConnectors = T,
    widthConnectors = 0.6,
    colConnectors = 'black',
    max.overlaps = 20,
    maxoverlapsConnectors = Inf)
SSc_lung_fibrosis
png(file='D:/Data/Data_drive/Data/IS_Protein_data/Tidied_scripts/plots/SSc_lung_fibrosis_volcano.png', width = 3000, height = 2000, res = 300)
SSc_lung_fibrosis

```
```{r}
# Required libraries
library(broom)
library(dplyr)

# Function to calculate R-squared, correlation, and p-value for each gene (SeqId)
calc_r2_pvalue_correlation <- function(data) {
  
  # Initialize an empty list to store results for each gene
  results <- list()
  
  # Iterate over each unique gene (SeqId)
  for (gene in unique(data$SeqId)) {
    
    # Subset data for the current gene
    gene_data <- data %>% filter(SeqId == gene)
    
    # Fit a linear model using Intensity as predictor and Lung_Fibrosis_binary as response
    model <- lm(Lung_Fibrosis_binary ~ Intensity, data = gene_data)
    
    # Calculate Pearson correlation coefficient
    correlation <- cor(gene_data$Intensity, gene_data$Lung_Fibrosis_binary, method = "pearson")
    
    # Extract R-squared and p-value from the model
    tidy_model <- tidy(model)  # For p-value
    glance_model <- glance(model)  # For R-squared
    
    # Store results in a list
    results[[gene]] <- data.frame(
      SeqId = gene,
      Correlation = correlation,  # Pearson correlation coefficient (r)
      R_squared = glance_model$r.squared,
      P_value = tidy_model$p.value[2]  # p-value for the Intensity predictor
    )
  }
  
  # Combine all results into a single data frame
  results_df <- do.call(rbind, results)
  
  return(results_df)
}

# Example usage:
subset_proteins_plot <- proteins_plot %>% filter(SampleGroup =="SSC")
results_table <- calc_r2_pvalue_correlation(subset_proteins_plot)

# Print or plot the results
 
results_table <- results_table %>% filter(P_value <0.05)
correlation_dataframe <- results_table %>% left_join(protein_IDs, by = "SeqId") %>%  select(Correlation, R_squared, P_value, EntrezGeneSymbol, SeqId)
```

```{r}
positive_correlations <- correlation_dataframe %>% filter(Correlation > 0) %>%  arrange(desc(R_squared)) %>% slice(1:10)
negative_correlations <- correlation_dataframe %>% filter(Correlation < 0) %>%  arrange(desc(R_squared)) %>% slice(1:10)
correlation_table <- full_join(positive_correlations,negative_correlations)
write.csv(correlation_table, "D:/Data/Data_drive/Data/IS_Protein_data/correlative_proteins_lung_fibrosis.csv")
```

```{r}
library(ggpubr)
library(ggpmisc)
library(cowplot)
mat_colors <- c('turquoise2','red')

SCUBE3 <- subset_proteins_plot  %>%
  subset(SeqId == "16773-29") %>%  
  ggplot(aes(Lung_Fibrosis, Intensity, label = SampleId))+
  facet_wrap( ~ EntrezGeneSymbol)+
  scale_y_log10()+
  geom_boxplot(fill = mat_colors)+
  theme_bw() +
  theme(legend.position = "top")+
  labs(color = "Sample", x="Lung fibrosis")
SCUBE3
ELP1 <- subset_proteins_plot  %>%
  subset(SeqId == "25102-23") %>%  
  ggplot(aes(Lung_Fibrosis, Intensity, label = SampleId))+
  facet_wrap( ~ EntrezGeneSymbol)+
  scale_y_log10()+
  geom_boxplot(fill = mat_colors)+
  theme_bw() +
  theme(legend.position = "top")+
  labs(color = "Sample", x="Lung fibrosis")
ELP1

SNTA1 <- subset_proteins_plot  %>%
  subset(SeqId == "22946-55") %>%  
  ggplot(aes(Lung_Fibrosis, Intensity, label = SampleId))+
  facet_wrap( ~ EntrezGeneSymbol)+
  scale_y_log10()+
  geom_boxplot(fill = mat_colors)+
  theme_bw() +
  theme(legend.position = "top")+
  labs(color = "Sample", x="Lung fibrosis")
SNTA1
	
HEXB <- subset_proteins_plot  %>%
  subset(SeqId == "6075-61") %>%  
  ggplot(aes(Lung_Fibrosis, Intensity, label = SampleId))+
  facet_wrap( ~ EntrezGeneSymbol)+
  scale_y_log10()+
  geom_boxplot(fill = mat_colors)+
  theme_bw() +
  theme(legend.position = "top")+
  labs(color = "Sample", x="Lung fibrosis")
HEXB
	
plot_grid(SCUBE3,ELP1,SNTA1,HEXB,ncol=4,nrow = 1)

```

```{r}
### Immunosuppression
subset_metadata <- dplyr::filter(metadata, grepl("Scleroderma", category))
subset_proteins <- normalised_expression[, colnames(normalised_expression) %in% rownames(subset_metadata)]
design <- model.matrix(~ Immunosupression_bin + 0, data = subset_metadata)
colnames(design) <- c("No","Yes")
contrastNames <- c("Yes - No")
contrasts <- makeContrasts(contrasts = contrastNames, levels=design)
limmaRes <- subset_proteins %>% lmFit(design = design)%>% contrasts.fit(contrasts) %>% eBayes()
coeffs <- coefficients(limmaRes)

stats_df <- topTable(limmaRes, number = nrow(subset_proteins))
stats_df <- merge(stats_df, protein_IDs, by = "row.names")
write.csv(stats_df, "D:/Data/Data_drive/Data/IS_Protein_data/SSc_immunosupressant_allproteins.csv")
writeable_SSc_immunosupression <- stats_df %>% filter( P.Value <0.05) %>% filter(logFC >0.585| logFC < -0.585) %>% select(SeqId, TargetFullName,Target,UniProt,EntrezGeneSymbol,logFC,AveExpr,P.Value)
write.csv(writeable_SSc_immunosupression, "D:/Data/Data_drive/Data/IS_Protein_data/SSc_immunosupressant.csv")

stats_df <- stats_df %>%
  arrange(P.Value) %>%
  mutate(EntrezGeneSymbol = as.character(EntrezGeneSymbol),
         label = ifelse(P.Value < 0.05 & EntrezGeneSymbol %in% head(EntrezGeneSymbol, 50), EntrezGeneSymbol, NA))

writeable_SSc_immunosupression %>%
  summarise(
    logFC_greater_than_0 = sum(logFC > 0),
    logFC_less_than_0 = sum(logFC < 0))

SSc_immunosupression <- EnhancedVolcano(data.frame(stats_df), x = 'logFC', y = 'P.Value',lab = stats_df$label,selectLab = stats_df$label,
    title = 'SSc Immunosupression vs SSc no Immunosupression',
    pCutoff = 0.05,
    FCcutoff = 0.585,
    xlim = c(min(stats_df[['logFC']], na.rm = TRUE), max(stats_df[['logFC']], na.rm = TRUE)),
    ylim = c(0, max(-log10(stats_df[['P.Value']]), na.rm = TRUE)),
    pointSize = 2.0,
    labSize = 4.0,
    labCol = 'grey14',
    colAlpha = 4/5,
    boxedLabels = T,
    legendPosition = 'None',
    legendLabSize = 8,
    legendIconSize = 3.0,
    drawConnectors = T,
    widthConnectors = 0.6,
    colConnectors = 'black',
    max.overlaps = 20,
    maxoverlapsConnectors = Inf)
SSc_immunosupression
png(file='D:/Data/Data_drive/Data/IS_Protein_data/Tidied_scripts/plots/SSc_immunosupressant_volcano.png', width = 3000, height = 2000, res = 300)
SSc_immunosupression

```


```{r}
subset_proteins_plot <- proteins_plot  %>%  dplyr::filter(grepl("SSC", SampleGroup))

CFDP1 <- subset_proteins_plot %>% 
  subset(SeqId == "24706-73") %>%  
  ggplot(aes(Immunosupression_bin, Intensity, label = SampleId))+
  facet_wrap( ~ EntrezGeneSymbol)+
  scale_y_log10()+
  geom_boxplot(fill = mat_colors)+
  theme_bw() +
  theme(legend.position = "top")+
  labs(color = "Sample", x="Immunosupression")
CFDP1

KRT1 <- subset_proteins_plot %>% 
  subset(SeqId == "9931-20") %>%  
  ggplot(aes(Immunosupression_bin, Intensity, label = SampleId))+
  facet_wrap( ~ EntrezGeneSymbol)+
  scale_y_log10()+
  geom_boxplot(fill = mat_colors)+
  theme_bw() +
  theme(legend.position = "top")+
  labs(color = "Sample", x="Immunosupression")
KRT1
```


```{r}
# Required libraries
library(broom)
library(dplyr)

# Function to calculate R-squared, correlation, and p-value for each gene (SeqId)
calc_r2_pvalue_correlation <- function(data) {
  
  # Initialize an empty list to store results for each gene
  results <- list()
  
  # Iterate over each unique gene (SeqId)
  for (gene in unique(data$SeqId)) {
    
    # Subset data for the current gene
    gene_data <- data %>% filter(SeqId == gene)
    
    # Fit a linear model using Intensity as predictor and Local_skin_score as response
    model <- lm(Local_skin_score ~ Intensity, data = gene_data)
    
    # Calculate Pearson correlation coefficient
    correlation <- cor(gene_data$Intensity, gene_data$Local_skin_score, method = "pearson")
    
    # Extract R-squared and p-value from the model
    tidy_model <- tidy(model)  # For p-value
    glance_model <- glance(model)  # For R-squared
    
    # Store results in a list
    results[[gene]] <- data.frame(
      SeqId = gene,
      Correlation = correlation,  # Pearson correlation coefficient (r)
      R_squared = glance_model$r.squared,
      P_value = tidy_model$p.value[2]  # p-value for the Intensity predictor
    )
  }
  
  # Combine all results into a single data frame
  results_df <- do.call(rbind, results)
  
  return(results_df)
}

# Example usage:
subset_proteins_plot <- proteins_plot %>% filter(SampleGroup =="SSC")
results_table <- calc_r2_pvalue_correlation(subset_proteins_plot)

# Print or plot the results
 
results_table <- results_table %>% filter(P_value <0.05)
correlation_dataframe <- results_table %>% left_join(protein_IDs, by = "SeqId") %>%  select(Correlation, R_squared, P_value, EntrezGeneSymbol, SeqId)
```

```{r}
positive_correlations <- correlation_dataframe %>% filter(Correlation > 0) %>%  arrange(desc(R_squared)) %>% slice(1:10)
negative_correlations <- correlation_dataframe %>% filter(Correlation < 0) %>%  arrange(desc(R_squared)) %>% slice(1:10)
correlation_table <- full_join(positive_correlations,negative_correlations)
write.csv(correlation_table, "D:/Data/Data_drive/Data/IS_Protein_data/correlative_proteins_local_skin_score.csv")
```

```{r}
library(ggpubr)
library(ggpmisc)
library(cowplot)

CHI3L1 <- subset_proteins_plot %>% 
  subset(SeqId =="11104-13")%>% 
  ggplot(aes(Local_skin_score, Intensity, label = SampleDescription))+
  facet_wrap( ~ EntrezGeneSymbol)+
   geom_point()+
    geom_label(color = "black", show.legend = FALSE) +
  stat_poly_line() +
  stat_poly_eq(use_label(c("R2", "p")), p.digits = 5) +
  theme(legend.position = "none")+
  labs( y = "",x="Local mRSS")
CHI3L1
TPSAB1 <- subset_proteins_plot %>% 
  subset(SeqId =="9409-11")%>% 
  ggplot(aes(Local_skin_score, Intensity, label = SampleDescription))+
  facet_wrap( ~ EntrezGeneSymbol)+
   geom_point()+
    geom_label(color = "black", show.legend = FALSE) +
  stat_poly_line() +
  stat_poly_eq(use_label(c("R2", "p")), p.digits = 5) +
  theme(legend.position = "none")+
  labs( y = "",x="Local mRSS")
TPSAB1
plot_grid(CHI3L1,TPSAB1,ncol=2,nrow = 1)

```


```{r}
# Required libraries
library(broom)
library(dplyr)

# Function to calculate R-squared, correlation, and p-value for each gene (SeqId)
calc_r2_pvalue_correlation <- function(data) {
  
  # Initialize an empty list to store results for each gene
  results <- list()
  
  # Iterate over each unique gene (SeqId)
  for (gene in unique(data$SeqId)) {
    
    # Subset data for the current gene
    gene_data <- data %>% filter(SeqId == gene)
    
    # Fit a linear model using Intensity as predictor and Disease_duration as response
    model <- lm(Disease_duration ~ Intensity, data = gene_data)
    
    # Calculate Pearson correlation coefficient
    correlation <- cor(gene_data$Intensity, gene_data$Disease_duration, method = "pearson")
    
    # Extract R-squared and p-value from the model
    tidy_model <- tidy(model)  # For p-value
    glance_model <- glance(model)  # For R-squared
    
    # Store results in a list
    results[[gene]] <- data.frame(
      SeqId = gene,
      Correlation = correlation,  # Pearson correlation coefficient (r)
      R_squared = glance_model$r.squared,
      P_value = tidy_model$p.value[2]  # p-value for the Intensity predictor
    )
  }
  
  # Combine all results into a single data frame
  results_df <- do.call(rbind, results)
  
  return(results_df)
}

# Example usage:
subset_proteins_plot <- proteins_plot %>% filter(SampleGroup =="SSC")
results_table <- calc_r2_pvalue_correlation(subset_proteins_plot)

# Print or plot the results
 
results_table <- results_table %>% filter(P_value <0.05)
correlation_dataframe <- results_table %>% left_join(protein_IDs, by = "SeqId") %>%  select(Correlation, R_squared, P_value, EntrezGeneSymbol, SeqId)
```

```{r}
positive_correlations <- correlation_dataframe %>% filter(Correlation > 0) %>%  arrange(desc(R_squared)) %>% slice(1:10)
negative_correlations <- correlation_dataframe %>% filter(Correlation < 0) %>%  arrange(desc(R_squared)) %>% slice(1:10)
correlation_table <- full_join(positive_correlations,negative_correlations)
write.csv(correlation_table, "D:/Data/Data_drive/Data/IS_Protein_data/correlative_proteins_disease_duration.csv")
```

```{r}
GPX1 <- subset_proteins_plot %>% 
  subset(SeqId =="15591-28")%>% 
  ggplot(aes(Disease_duration, Intensity, label = SampleDescription))+
  facet_wrap( ~ EntrezGeneSymbol)+
   geom_point()+
    geom_label(color = "black", show.legend = FALSE) +
  stat_poly_line() +
  stat_poly_eq(use_label(c("R2", "p")), p.digits = 5) +
  theme(legend.position = "none")+
  labs( y = "",x="Disease Duration (Months)")
GPX1

PPP1R9B <- subset_proteins_plot %>% 
  subset(SeqId =="21991-79")%>% 
  ggplot(aes(Disease_duration, Intensity, label = SampleDescription))+
  facet_wrap( ~ EntrezGeneSymbol)+
   geom_point()+
    geom_label(color = "black", show.legend = FALSE) +
  stat_poly_line() +
  stat_poly_eq(use_label(c("R2", "p")), p.digits = 5) +
  theme(legend.position = "none")+
  labs( y = "",x="Disease Duration (Months)")
PPP1R9B
plot_grid(GPX1,PPP1R9B,ncol=2,nrow = 1)

```