#Code for manuscript entitled: "Immunoglobulin signature predicts risk of post-acute COVID-19 syndrome", Cervia et al., Nature Communications, 2021

# Legend:

# 1. Import packages and data
# 2. Data wrangling and grouping
# 3. Functions for data analysis
# 4. Analysis of total immunoglobulin levels
#   a) IgM
#   b) IgG1
#   c) IgG3
#   d) IgA
#   e) IgG2
#   f) IgG4
# 5. Development of prediction model - Variables
#   a) Linearity of variables
#   b) Age
#   c) Number of symptoms during primary infection
#   d) History of asthma bronchiale
#   e) COVID-19 severity
# 6. Development of prediction model - PACS score
#   a) Primary infection
#   b) 6-months follow-up
#   c) Decision curve analysis
#   d) Interaction plots
# 7. Code for implementation of PACS score




#Import packages and data----
library("grid")
#install.packages("biostatUZH", repos = "http://R-Forge.R-project.org") 
library("biostatUZH")                 #version 1.8.0
# install.packages("remotes")
# remotes::install_github("BavoDC/CalibrationCurves")
library("CalibrationCurves")          #version 0.1.2
library("ggplot2")                    #version 3.3.5
library("ggpubr")                     #version 0.4.0
library("tidyverse")                  #version 1.3.1
library("readxl")                     #version 1.3.1
library("rstudioapi")                 #version 0.13
library("epitools")                   #version 0.5-10.1
library("pROC")                       #version 1.18.0
library("gbm")                        #version 2.1.8
library("shrink")                     #version 1.2.1
library("ggstatsplot")                #version 0.9.0
library("dcurves")                    #version 0.2.0
library("mgcv")                       #version 1.8-38
library("sjPlot")                     #version 2.8.9 
library("scales")                     #version 1.1.1
library("interactions")               #version 1.1.5


#Data wrangling and grouping----

#Import Data
current_path = rstudioapi::getActiveDocumentContext()$path
setwd(dirname(current_path ))

xdata <- read_excel("Data_Cervia_et_al_NatComm_2021_v2.xlsx", na = "NA")  #Enter excel file name

#Factors

xdata$Sampling_month <- factor(xdata$Sampling_month,
                               levels = c("ONE","SIX", "TWELVE"))

xdata$COVID <- factor(xdata$COVID,
                      levels = c("Healthy","COVID"))

xdata$SevMax2 <- factor(xdata$SevMax2,
                        levels = c("Healthy","Mild","Severe"))

xdata$SevMax6 <- factor(xdata$SevMax6,
                        levels = c("Healthy", "Asymptomatic","Mild","Pneu" ,"Sev Pneu" ,"Mild ARDS" ,"Mod ARDS", "Sev ARDS"))

xdata$Hospitalized <- factor(xdata$Hospitalized,
                             levels = c("NO","YES"))

xdata$PACS <- factor(xdata$PACS,
                      levels = c("NO","YES"))


#Add Healhty  to six month measurements
healthy<-xdata%>%filter(SevMax2=="Healthy")%>%mutate(Sampling_month="SIX", Complete_OST = "NO")
xdata<-rbind(xdata,healthy)



#Create groups
xdata_selectR <- xdata[xdata$Resampling %in% c("YES"), ] #followed-up patients
xdata_selectR2 <- xdata[xdata$Resampling_andH %in% c("YES"), ]  #followed-up patients + healthy controls

xdata_selectONE <- xdata[xdata$Sampling_month %in% c("ONE"), ]  #Primary infection
xdata_selectRONE <- xdata_selectR[xdata_selectR$Sampling_month %in% c("ONE"), ] 
xdata_selectR2ONE <- xdata_selectR2[xdata_selectR2$Sampling_month %in% c("ONE"), ]

xdata_selectRSIX <- xdata_selectR[xdata_selectR$Sampling_month %in% c("SIX"), ] #6-month follow-up
xdata_selectR2SIX <- xdata_selectR2[xdata_selectR2$Sampling_month %in% c("SIX"), ]

xdata_selectR2OS <- xdata_selectR2[xdata_selectR2$Sampling_month %in% c("ONE", "SIX"), ] #Primary infection & 6-month follow-up

xdata_selectRnH <- xdata_selectR[xdata_selectR$SevMax2 %in% c("Mild", "Severe"), ]          #Without Healthy controls
xdata_selectRONEnH <- xdata_selectRONE[xdata_selectRONE$SevMax2 %in% c("Mild", "Severe"), ]
xdata_selectRSIXnH <- xdata_selectRSIX[xdata_selectRSIX$SevMax2 %in% c("Mild", "Severe"), ]
xdata_selectRnHOS <- xdata_selectRnH[xdata_selectRnH$Sampling_month %in% c("ONE", "SIX"), ]


xdata_selectONEH <- xdata_selectONE[xdata_selectONE$SevMax2 %in% c( "Healthy"), ]   #Only Healhty controls

xdata_selectRONE_Asthma <- xdata_selectRONE[xdata_selectRONE$Asthma %in% c("YES"), ]  #Patients with history of asthma
xdata_selectRONE_nAsthma <- xdata_selectRONE[xdata_selectRONE$Asthma %in% c("NO"), ]

xdata_selectRONEHosp <- xdata_selectRONE[xdata_selectRONE$Hospitalized %in% c( "YES"), ]  #Hospitalized patients
xdata_selectRONEnHosp <- xdata_selectRONE[xdata_selectRONE$Hospitalized %in% c( "NO"), ]

xdata_selectR2ONEPACS <- xdata_selectR2ONE[xdata_selectR2ONE$PACS %in% c("YES"), ]  #Patients with PACS
xdata_selectR2ONEnPACS <- xdata_selectR2ONE[xdata_selectR2ONE$PACS %in% c("NO"), ]

xdata_selectRcOST <- xdata_selectR[xdata_selectR$Complete_OST %in% c("YES"), ]  #Complete observaions: Patients with blood sampling at all three patient visits
xdata_selectR2cOST <- xdata_selectR2[xdata_selectR2$Complete_OST %in% c("YES"), ]
xdata_selectRcOSTPACS <- xdata_selectRcOST[xdata_selectRcOST$PACS %in% c("YES"), ]
xdata_selectRcOSTnPACS <- xdata_selectRcOST[xdata_selectRcOST$PACS %in% c("NO"), ]

#Comparisons and color vectors
my_comparisons<-list(c("Healthy","Mild") ,c("Healthy","Severe"),c("Mild","Severe"))
my_colors_PACS <- c("#ffa44a","#c7264f")

#Functions----
alpha <- scales::alpha

GeomSplitViolin <- ggproto("GeomSplitViolin", GeomViolin,
                           draw_group = function(self, data, ..., draw_quantiles = NULL) {
                             # Original function by Jan Gleixner (@jan-glx)
                             # Adjustments by Wouter van der Bijl (@Axeman)
                             data <- transform(data, xminv = x - violinwidth * (x - xmin), xmaxv = x + violinwidth * (xmax - x))
                             grp <- data[1, "group"]
                             newdata <- plyr::arrange(transform(data, x = if (grp %% 2 == 1) xminv else xmaxv), if (grp %% 2 == 1) y else -y)
                             newdata <- rbind(newdata[1, ], newdata, newdata[nrow(newdata), ], newdata[1, ])
                             newdata[c(1, nrow(newdata) - 1, nrow(newdata)), "x"] <- round(newdata[1, "x"])
                             if (length(draw_quantiles) > 0 & !scales::zero_range(range(data$y))) {
                               stopifnot(all(draw_quantiles >= 0), all(draw_quantiles <= 1))
                               quantiles <- create_quantile_segment_frame(data, draw_quantiles, split = TRUE, grp = grp)
                               aesthetics <- data[rep(1, nrow(quantiles)), setdiff(names(data), c("x", "y")), drop = FALSE]
                               aesthetics$alpha <- rep(1, nrow(quantiles))
                               both <- cbind(quantiles, aesthetics)
                               quantile_grob <- GeomPath$draw_panel(both, ...)
                               ggplot2:::ggname("geom_split_violin", grid::grobTree(GeomPolygon$draw_panel(newdata, ...), quantile_grob))
                             }
                             else {
                               ggplot2:::ggname("geom_split_violin", GeomPolygon$draw_panel(newdata, ...))
                             }
                           }
)

create_quantile_segment_frame <- function(data, draw_quantiles, split = FALSE, grp = NULL) {
  dens <- cumsum(data$density) / sum(data$density)
  ecdf <- stats::approxfun(dens, data$y)
  ys <- ecdf(draw_quantiles)
  violin.xminvs <- (stats::approxfun(data$y, data$xminv))(ys)
  violin.xmaxvs <- (stats::approxfun(data$y, data$xmaxv))(ys)
  violin.xs <- (stats::approxfun(data$y, data$x))(ys)
  if (grp %% 2 == 0) {
    data.frame(
      x = ggplot2:::interleave(violin.xs, violin.xmaxvs),
      y = rep(ys, each = 2), group = rep(ys, each = 2)
    )
  } else {
    data.frame(
      x = ggplot2:::interleave(violin.xminvs, violin.xs),
      y = rep(ys, each = 2), group = rep(ys, each = 2)
    )
  }
}

geom_split_violin <- function(mapping = NULL, data = NULL, stat = "ydensity", position = "identity", ..., 
                              draw_quantiles = NULL, trim = TRUE, scale = "area", na.rm = FALSE, 
                              show.legend = NA, inherit.aes = TRUE) {
  layer(data = data, mapping = mapping, stat = stat, geom = GeomSplitViolin, position = position, 
        show.legend = show.legend, inherit.aes = inherit.aes, 
        params = list(trim = trim, scale = scale, draw_quantiles = draw_quantiles, na.rm = na.rm, ...))
}

dodge <- position_dodge(width=.5)



#Total immunoglobulins----


#__IgM----
ggplot(xdata_selectR2OS, aes(x=Sampling_month, y=IgM, fill=COVID))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = COVID), label = "p.format")+
  labs(x="Sampling month", fill="", y="Total IgM (g/l)")+
  scale_fill_manual(values=alpha(c("Chartreuse3","#495363"),0.8))+
  ylim(-0.4,5.2) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2ONE, aes(x=SevMax2, y=IgM, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons, method="wilcox.test", label.y = c(4.6,4.79,4.99))+
  labs(x="", fill="", y="Total IgM (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  ylim(-0.4,5.2) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2SIX, aes(x=SevMax2, y=IgM, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons, method="wilcox.test")+
  labs(x="", fill="", y="Total IgM (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E")) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectRONEnH, aes(x=Age, y=IgM)) +
  theme_classic()+
  geom_point(size = 3.5, colour = "#495363") +
  theme(text = element_text(size=15), 
        axis.text.x = element_text(size=15), 
        axis.title.x = element_text(size=15),
        plot.title = element_text(hjust = 0.5)) + 
  geom_smooth(method="lm", col="black", se = T, size =2, alpha=0.3)+
  labs(x="Patient age (years)", y="Total IgM (g/l)", colour= "")

CAlm3_IgM <- lm(IgM ~ Age, data = xdata_selectRONEnH)
CAlm3_IgM_s <- coef(summary(CAlm3_IgM))
CAlm3_IgM_p <- CAlm3_IgM_s["Age","Pr(>|t|)"]
grid.text(paste("p =", biostatUZH::formatPval(CAlm3_IgM_p)), just = "left", x=0.2,  y=0.95, gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("R2adj =", round(summary(CAlm3_IgM)$adj.r.squared, 3)), just = "left", x=0.3,  y=0.95, gp=gpar(fontsize=11.5, col="black"))


ggplot(xdata_selectRnHOS, aes(x=Sampling_month, y=IgM, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format")+
  labs(x="Sampling month", fill="PACS", y="Total IgM (g/l)")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

M_IgM<- median(na.omit(xdata_selectONEH$IgM))

ggplot(data = xdata_selectR2cOST, aes(x = Days, y = IgM, group = PatID))+
  geom_point(aes(shape = PACS, color=COVID), size=3.5)+
  scale_shape_manual(values=c( 1, 16)) +
  scale_color_manual(values=c( "chartreuse3", "black")) +
  geom_line() +
  theme_classic() +
  stat_smooth(data=xdata_selectRcOSTnPACS ,group=1, method="gam", size=1.5, col="#ffa44a", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  stat_smooth(data=xdata_selectRcOSTPACS ,group=1, method="gam", size=1.5, col= "#c7264f", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  labs(x="Time after symptom onset (days)", y="Total IgM (g/l)", shape= "PACS", colour="")+
  geom_segment(aes(x = -60, y = M_IgM, xend = -40, yend = M_IgM), size=1.5, color="chartreuse3")

gam0 <- gam(IgM ~ s(Days, k=3), method="REML", data = xdata_selectRcOST, trace = TRUE)
gam0.s <- summary(gam0)
gam0.p <- gam0.s$s.table[,"p-value"]
grid.text(paste("R2adj =", round(gam0.s$r.sq, 2), ", p =", biostatUZH::formatPval(gam0.p)), x=0.6,  y=0.927, gp=gpar(fontsize=11.5))



#__IgG1----
ggplot(xdata_selectR2OS, aes(x=Sampling_month, y=IgG1, fill=COVID))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = COVID), label = "p.format")+
  labs(x="Sampling month", fill="", y="Total IgG1 (g/l)")+
  scale_fill_manual(values=alpha(c("Chartreuse3","#495363"),0.8))+
  ylim(-0.4,15.5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


ggplot(xdata_selectR2ONE, aes(x=SevMax2, y=IgG1, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons, method="wilcox.test", label.y = c(12,14,14.9))+
  labs(x="", fill="", y="Total IgG1 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  ylim(-0.4,15.5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2SIX, aes(x=SevMax2, y=IgG1, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="", fill="", y="Total IgG1 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E")) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectRONEnH, aes(x=Age, y=IgG1)) +
  theme_classic()+
  geom_point(size = 3.5, colour = "#495363") +
  theme(text = element_text(size=15), 
        axis.text.x = element_text(size=15), 
        axis.title.x = element_text(size=15),
        plot.title = element_text(hjust = 0.5)) + 
  geom_smooth(method="lm", col="black", se = T, size =2, alpha=0.3)+
  labs(x="Patient age (years)", y="Total IgG1 (g/l)", colour= "")

CAlm3_IgG1 <- lm(IgG1 ~ Age, data = xdata_selectRONEnH)
CAlm3_IgG1_s <- coef(summary(CAlm3_IgG1))
CAlm3_IgG1_p <- CAlm3_IgG1_s["Age","Pr(>|t|)"]
grid.text(paste("p =", biostatUZH::formatPval(CAlm3_IgG1_p)), just = "left", x=0.2,  y=0.95, gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("R2adj =", round(summary(CAlm3_IgG1)$adj.r.squared, 3)), just = "left", x=0.3,  y=0.95, gp=gpar(fontsize=11.5, col="black"))


ggplot(xdata_selectRnHOS, aes(x=Sampling_month, y=IgG1, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles =  0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format")+
  labs(x="Sampling month", fill="PACS", y="Total IgG1 (g/l)")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


M_IgG1<- median(na.omit(xdata_selectONEH$IgG1))

ggplot(data = xdata_selectR2cOST, aes(x = Days, y = IgG1, group = PatID))+
  geom_point(aes(shape = PACS, color=COVID), size=3.5)+
  scale_shape_manual(values=c( 1, 16)) +
  scale_color_manual(values=c( "chartreuse3", "black")) +
  geom_line() +
  theme_classic() +
  stat_smooth(data=xdata_selectRcOSTnPACS ,group=1, method="gam", size=1.5, col="#ffa44a", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  stat_smooth(data=xdata_selectRcOSTPACS ,group=1, method="gam", size=1.5, col= "#c7264f", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  labs(x="Time after symptom onset (days)", y="Total IgG1 (g/l)", shape= "PACS", colour="")+
  geom_segment(aes(x = -60, y = M_IgG1, xend = -40, yend = M_IgG1), size=1.5, color="chartreuse3")

gam1 <- gam(IgG1 ~ s(Days, k=3), method="REML", data = xdata_selectRcOST, trace = TRUE)
gam1.s <- summary(gam1)
gam1.p <- gam1.s$s.table[,"p-value"]
grid.text(paste("R2adj =", round(gam1.s$r.sq, 2), ", p =", biostatUZH::formatPval(gam1.p)), x=0.6,  y=0.927, gp=gpar(fontsize=11.5))



#IgG1_per
ggplot(xdata_selectR2ONEnPACS, aes(x=SevMax2, y=IgG1_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="Patients without PACS", fill="", y="Total IgG1 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2ONEPACS, aes(x=SevMax2, y=IgG1_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means()+
  labs(x="Patients with PACS", fill="", y="Total IgG1 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


#__IgG3----
ggplot(xdata_selectR2OS, aes(x=Sampling_month, y=IgG3, fill=COVID))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = COVID), label = "p.format")+
  labs(x="Sampling month", fill="", y="Total IgG3 (g/l)")+
  scale_fill_manual(values=alpha(c("Chartreuse3","#495363"),0.8))+
  ylim(-0.4,4.5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2ONE, aes(x=SevMax2, y=IgG3, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons, method="wilcox.test", label.y = c(3,4.4,4.8))+
  labs(x="", fill="", y="Total IgG3 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  ylim(-0.4,5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2SIX, aes(x=SevMax2, y=IgG3, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="", fill="", y="Total IgG3 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E")) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectRONEnH, aes(x=Age, y=IgG3)) +
  theme_classic()+
  geom_point(size = 3.5, colour = "#495363") +
  theme(text = element_text(size=15), 
        axis.text.x = element_text(size=15), 
        axis.title.x = element_text(size=15),
        plot.title = element_text(hjust = 0.5)) + 
  geom_smooth(method="lm", col="black", se = T, size =2, alpha=0.3)+
  labs(x="Patient age (years)", y="Total IgG3 (g/l)", colour= "")

CAlm3_IgG3 <- lm(IgG3 ~ Age, data = xdata_selectRONEnH)
CAlm3_IgG3_s <- coef(summary(CAlm3_IgG3))
CAlm3_IgG3_p <- CAlm3_IgG3_s["Age","Pr(>|t|)"]
grid.text(paste("p =", biostatUZH::formatPval(CAlm3_IgG3_p)), just = "left", x=0.2,  y=0.95, gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("R2adj =", round(summary(CAlm3_IgG3)$adj.r.squared, 3)), just = "left", x=0.3,  y=0.95, gp=gpar(fontsize=11.5, col="black"))


ggplot(xdata_selectRnHOS, aes(x=Sampling_month, y=IgG3, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format")+
  labs(x="Sampling month", fill="PACS", y="Total IgG3 (g/l)")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


M_IgG3<- median(na.omit(xdata_selectONEH$IgG3))

ggplot(data = xdata_selectR2cOST, aes(x = Days, y = IgG3, group = PatID))+
  geom_point(aes(shape = PACS, color=COVID), size=3.5)+
  scale_shape_manual(values=c( 1, 16)) +
  scale_color_manual(values=c( "chartreuse3", "black")) +
  geom_line() +
  theme_classic() +
  stat_smooth(data=xdata_selectRcOSTnPACS ,group=1, method="gam", size=1.5, col="#ffa44a", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  stat_smooth(data=xdata_selectRcOSTPACS ,group=1, method="gam", size=1.5, col= "#c7264f", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  labs(x="Time after symptom onset (days)", y="Total IgG3 (g/l)", shape= "PACS", colour="")+
  geom_segment(aes(x = -60, y = M_IgG3, xend = -40, yend = M_IgG3), size=1.5, color="chartreuse3")

gam2 <- gam(IgG3 ~ s(Days, k=3), method="REML", data = xdata_selectRcOST, trace = TRUE)
gam2.s <- summary(gam2)
gam2.p <- gam2.s$s.table[,"p-value"]
grid.text(paste("R2adj =", round(gam2.s$r.sq, 2), ", p =", biostatUZH::formatPval(gam2.p)), x=0.6,  y=0.927, gp=gpar(fontsize=11.5))


#IgG3_per
ggplot(xdata_selectR2ONEnPACS, aes(x=SevMax2, y=IgG3_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="Patients without PACS", fill="", y="Total IgG3 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2ONEPACS, aes(x=SevMax2, y=IgG3_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means()+
  labs(x="Patients with PACS", fill="", y="Total IgG3 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

#__IgA----
ggplot(xdata_selectR2OS, aes(x=Sampling_month, y=IgA, fill=COVID))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = COVID), label = "p.format")+
  labs(x="Sampling month", fill="", y="Total IgA (g/l)")+
  scale_fill_manual(values=alpha(c("Chartreuse3","#495363"),0.8))+
  ylim(-0.4,6.5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


ggplot(xdata_selectR2ONE, aes(x=SevMax2, y=IgA, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons, method="wilcox.test", label.y = c(5,5.6,6.2))+
  labs(x="", fill="", y="Total IgA (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  ylim(-0.4,6.5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2SIX, aes(x=SevMax2, y=IgA, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="", fill="", y="Total IgA (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E")) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectRONEnH, aes(x=Age, y=IgA)) +
  theme_classic()+
  geom_point(size = 3.5, colour = "#495363") +
  theme(text = element_text(size=15), 
        axis.text.x = element_text(size=15), 
        axis.title.x = element_text(size=15),
        plot.title = element_text(hjust = 0.5)) + 
  geom_smooth(method="lm", col="black", se = T, size =2, alpha=0.3)+
  labs(x="Patient age (years)", y="Total IgA (g/l)", colour= "")

CAlm3_IgA <- lm(IgA ~ Age, data = xdata_selectRONEnH)
CAlm3_IgA_s <- coef(summary(CAlm3_IgA))
CAlm3_IgA_p <- CAlm3_IgA_s["Age","Pr(>|t|)"]
grid.text(paste("p =", biostatUZH::formatPval(CAlm3_IgA_p)), just = "left", x=0.2,  y=0.95, gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("R2adj =", round(summary(CAlm3_IgA)$adj.r.squared, 3)), just = "left", x=0.3,  y=0.95, gp=gpar(fontsize=11.5, col="black"))


ggplot(xdata_selectRnHOS, aes(x=Sampling_month, y=IgA, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format")+
  labs(x="Sampling month", fill="PACS", y="Total IgA (g/l)")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


M_IgA<- median(na.omit(xdata_selectONEH$IgA))

ggplot(data = xdata_selectR2cOST, aes(x = Days, y = IgA, group = PatID))+
  geom_point(aes(shape = PACS, color=COVID), size=3.5)+
  scale_shape_manual(values=c( 1, 16)) +
  scale_color_manual(values=c( "chartreuse3", "black")) +
  geom_line() +
  theme_classic() +
  stat_smooth(data=xdata_selectRcOSTnPACS ,group=1, method="gam", size=1.5, col="#ffa44a", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  stat_smooth(data=xdata_selectRcOSTPACS ,group=1, method="gam", size=1.5, col= "#c7264f", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  labs(x="Time after symptom onset (days)", y="Total IgA (g/l)", shape= "PACS", colour="")+
  geom_segment(aes(x = -60, y = M_IgA, xend = -40, yend = M_IgA), size=1.5, color="chartreuse3")

gam3 <- gam(IgA ~ s(Days, k=3), method="REML", data = xdata_selectRcOST, trace = TRUE)
gam3.s <- summary(gam3)
gam3.p <- gam3.s$s.table[,"p-value"]
grid.text(paste("R2adj =", round(gam3.s$r.sq, 2), ", p =", biostatUZH::formatPval(gam3.p)), x=0.6,  y=0.927, gp=gpar(fontsize=11.5))



#__IgG2----
ggplot(xdata_selectR2OS, aes(x=Sampling_month, y=IgG2, fill=COVID))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = COVID), label = "p.format")+
  labs(x="Sampling month", fill="", y="Total IgG2 (g/l)")+
  scale_fill_manual(values=alpha(c("Chartreuse3","#495363"),0.8))+
  ylim(-0.4,12) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


ggplot(xdata_selectR2ONE, aes(x=SevMax2, y=IgG2, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons, method="wilcox.test")+
  labs(x="", fill="", y="Total IgG2 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  ylim(-0.4,12) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2SIX, aes(x=SevMax2, y=IgG2, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="", fill="", y="Total IgG2 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E")) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectRONEnH, aes(x=Age, y=IgG2)) +
  theme_classic()+
  geom_point(size = 3.5, colour = "#495363") +
  theme(text = element_text(size=15), 
        axis.text.x = element_text(size=15), 
        axis.title.x = element_text(size=15),
        plot.title = element_text(hjust = 0.5)) + 
  geom_smooth(method="lm", col="black", se = T, size =2, alpha=0.3)+
  labs(x="Patient age (years)", y="Total IgG2 (g/l)", colour= "")

CAlm3_IgG2 <- lm(IgG2 ~ Age, data = xdata_selectRONEnH)
CAlm3_IgG2_s <- coef(summary(CAlm3_IgG2))
CAlm3_IgG2_p <- CAlm3_IgG2_s["Age","Pr(>|t|)"]
grid.text(paste("p =", biostatUZH::formatPval(CAlm3_IgG2_p)), just = "left", x=0.2,  y=0.95, gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("R2adj =", round(summary(CAlm3_IgG2)$adj.r.squared, 3)), just = "left", x=0.3,  y=0.95, gp=gpar(fontsize=11.5, col="black"))


ggplot(xdata_selectRnHOS, aes(x=Sampling_month, y=IgG2, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format")+
  labs(x="Sampling month", fill="PACS", y="Total IgG2 (g/l)")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

M_IgG2<- median(na.omit(xdata_selectONEH$IgG2))

ggplot(data = xdata_selectR2cOST, aes(x = Days, y = IgG2, group = PatID))+
  geom_point(aes(shape = PACS, color=COVID), size=3.5)+
  scale_shape_manual(values=c( 1, 16)) +
  scale_color_manual(values=c( "chartreuse3", "black")) +
  geom_line() +
  theme_classic() +
  stat_smooth(data=xdata_selectRcOSTnPACS ,group=1, method="gam", size=1.5, col="#ffa44a", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  stat_smooth(data=xdata_selectRcOSTPACS ,group=1, method="gam", size=1.5, col= "#c7264f", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  labs(x="Time after symptom onset (days)", y="Total IgG2 (g/l)", shape= "PACS", colour="")+
  geom_segment(aes(x = -60, y = M_IgG2, xend = -40, yend = M_IgG2), size=1.5, color="chartreuse3")

gam4 <- gam(IgG2 ~ s(Days, k=3), method="REML", data = xdata_selectRcOST, trace = TRUE)
gam4.s <- summary(gam4)
gam4.p <- gam4.s$s.table[,"p-value"]
grid.text(paste("R2adj =", round(gam4.s$r.sq, 2), ", p =", biostatUZH::formatPval(gam4.p)), x=0.6,  y=0.927, gp=gpar(fontsize=11.5))



#IgG2_per
ggplot(xdata_selectR2ONEnPACS, aes(x=SevMax2, y=IgG2_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="Patients without PACS", fill="", y="Total IgG2 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2ONEPACS, aes(x=SevMax2, y=IgG2_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means()+
  labs(x="Patients with PACS", fill="", y="Total IgG2 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

#__IgG4----
ggplot(xdata_selectR2OS, aes(x=Sampling_month, y=IgG4, fill=COVID))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = COVID), label = "p.format")+
  labs(x="Sampling month", fill="", y="Total IgG4 (g/l)")+
  scale_fill_manual(values=alpha(c("Chartreuse3","#495363"),0.8))+
  ylim(-0.4,3.5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


ggplot(xdata_selectR2ONE, aes(x=SevMax2, y=IgG4, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons, method="wilcox.test")+
  labs(x="", fill="", y="Total IgG4 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  ylim(-0.4,3.5) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2SIX, aes(x=SevMax2, y=IgG4, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="", fill="", y="Total IgG4 (g/l)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E")) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectRONEnH, aes(x=Age, y=IgG4)) +
  theme_classic()+
  geom_point(size = 3.5, colour = "#495363") +
  theme(text = element_text(size=15), 
        axis.text.x = element_text(size=15), 
        axis.title.x = element_text(size=15),
        plot.title = element_text(hjust = 0.5)) + 
  geom_smooth(method="lm", col="black", se = T, size =2, alpha=0.3)+
  labs(x="Patient age (years)", y="Total IgG4 (g/l)", colour= "")

CAlm3_IgG4 <- lm(IgG4 ~ Age, data = xdata_selectRONEnH)
CAlm3_IgG4_s <- coef(summary(CAlm3_IgG4))
CAlm3_IgG4_p <- CAlm3_IgG4_s["Age","Pr(>|t|)"]
grid.text(paste("p =", biostatUZH::formatPval(CAlm3_IgG4_p)), just = "left", x=0.2,  y=0.95, gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("R2adj =", round(summary(CAlm3_IgG4)$adj.r.squared, 3)), just = "left", x=0.3,  y=0.95, gp=gpar(fontsize=11.5, col="black"))


ggplot(xdata_selectRnHOS, aes(x=Sampling_month, y=IgG4, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format")+
  labs(x="Sampling month", fill="PACS", y="Total IgG4 (g/l)")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

M_IgG4<- median(na.omit(xdata_selectONEH$IgG4))

ggplot(data = xdata_selectR2cOST, aes(x = Days, y = IgG4, group = PatID))+
  geom_point(aes(shape = PACS, color=COVID), size=3.5)+
  scale_shape_manual(values=c( 1, 16)) +
  scale_color_manual(values=c( "chartreuse3", "black")) +
  geom_line() +
  theme_classic() +
  stat_smooth(data=xdata_selectRcOSTnPACS ,group=1, method="gam", size=1.5, col="#ffa44a", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  stat_smooth(data=xdata_selectRcOSTPACS ,group=1, method="gam", size=1.5, col= "#c7264f", se = TRUE, formula=y ~ s(x, bs = "cs", k=3))+
  labs(x="Time after symptom onset (days)", y="Total IgG4 (g/l)", shape= "PACS", colour="")+
  geom_segment(aes(x = -60, y = M_IgG4, xend = -40, yend = M_IgG4), size=1.5, color="chartreuse3")

gam5 <- gam(IgG4 ~ s(Days, k=3), method="REML", data = xdata_selectRcOST, trace = TRUE)
gam5.s <- summary(gam5)
gam5.p <- gam5.s$s.table[,"p-value"]
grid.text(paste("R2adj =", round(gam5.s$r.sq, 2), ", p =", biostatUZH::formatPval(gam5.p)), x=0.6,  y=0.927, gp=gpar(fontsize=11.5))



#IgG4_per
ggplot(xdata_selectR2ONEnPACS, aes(x=SevMax2, y=IgG4_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means(comparison=my_comparisons)+
  labs(x="Patients without PACS", fill="", y="Total IgG4 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)

ggplot(xdata_selectR2ONEPACS, aes(x=SevMax2, y=IgG4_per, fill=SevMax2))+
  geom_violin(trim = F, draw_quantiles = c(0.5))+
  geom_point(position=position_jitterdodge(),alpha=0.4, size=2.8)+
  theme_classic()+
  stat_compare_means()+
  labs(x="Patients with PACS", fill="", y="Total IgG4 (% of total IgG)")+
  scale_fill_manual(values=c("chartreuse3","#A6C6F0","#F58E8E"))+
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)



#Prediction model development

#Predictor variables----

#__Linearity of variables----
plot(gam(PACS ~ s(Age), family=binomial(link="logit"), data = xdata_selectRONEnH, trace = TRUE))
plot(gam(PACS ~ s(scale(Age)), family=binomial(link="logit"), data = xdata_selectRONEnH, trace = TRUE))

plot(gam(PACS ~ s(Nr_Symptoms, k=6), family=binomial(link="logit"), data = xdata_selectRONEnH, trace = TRUE))

plot(gam(PACS ~ s(IgM), family=binomial(link="logit"), data = xdata_selectRONEnH, trace = TRUE))

plot(gam(PACS ~ s(IgG3), family=binomial(link="logit"), data = xdata_selectRONEnH, trace = TRUE))

#__Age----
ggplot(xdata_selectRONEnH, aes(x=Sampling_month, y=Age, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format", label.y=120)+
  labs(x="Sampling month", fill="PACS", y="Patient age")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


#__Nr of symptoms----
ggplot(xdata_selectRONEnH, aes(x=Sampling_month, y=Nr_Symptoms, fill=PACS))+
  geom_split_violin(trim = F, draw_quantiles = 0.5)+
  geom_point(shape=16,position=dodge)+
  theme_classic()+
  stat_compare_means(aes(group = PACS), label = "p.format", label.y=6)+
  labs(x="Sampling month", fill="PACS", y="Number of symptoms (0-5)")+
  scale_fill_manual(values=my_colors_PACS) +
  geom_boxplot(width = 0.25, notch = FALSE, notchwidth = .4, outlier.shape = NA, coef=0)


ggplot(xdata_selectRONEnH, aes(x=SevMax6, y=Nr_Symptoms))+
  geom_boxplot()+
  geom_point(shape=16,position="dodge")+
  theme_classic()+
  stat_compare_means(aes(group = SevMax6), label.y=6)+
  labs(x="Disease severity",  y="Number of symptoms (0-5)")


Symp<- table(xdata_selectRONEnH$Nr_Symptoms, xdata_selectRONEnH$PACS)
Symp_per <-table(xdata_selectRONEnH$Nr_Symptoms, xdata_selectRONEnH$PACS)[,2] /  table(xdata_selectRONEnH$Nr_Symptoms)
barplot(t(Symp),main="",
        xlab="Number of symptoms during acute COVID-19", col=c("#ffa44a","#c7264f"),legend = c("NO PACS", "PACS"),)
barplot(Symp_per,main="",ylab="PACS (%)",
        xlab="Number of symptoms during acute COVID-19", col=c("#c7264f"))


#__Asthma----
pie(table(xdata_selectRONE_nAsthma$PACS), main="Patients without \n history of asthma bronchiale", init.angle = 90,
    clockwise = TRUE,  labels=c("Acute COVID-19", "Post-acute COVID-19 Syndrome"),
    col=c("#ffa44a","#c7264f"))

pie(table(xdata_selectRONE_Asthma$PACS), main="Patients with \n history of asthma bronchiale", init.angle = 90,
    clockwise = TRUE,  labels=c("Acute COVID-19", "Post-acute COVID-19 Syndrome"),
    col=c("#ffa44a","#c7264f"))

ggplot(xdata_selectONE, aes(x=Asthma, y=IgG3))+
  geom_boxplot(outlier.shape=NA,fill="#F58E8E")+  
  geom_point(position=position_jitterdodge(),alpha=0.8, aes(col=SevMax6), size=2.8)+
  stat_compare_means()+
  theme_classic()+
  scale_color_manual(values=c("chartreuse3","grey","#A6C6F0", "#4D77EB","#F58E8E", 
                              "#E64A4A", "#D91A1A", "#8F0B0B"))+
  labs(x="Asthma bronchiale", y="Total IgG3 (g/l)", colour= "COVID-19 severity")


#__COVID-19 Severity/Hospitalization----

Sev<- table(xdata_selectRONEnH$SevMax6, xdata_selectRONEnH$PACS)
Sev_per <-table(xdata_selectRONEnH$SevMax6, xdata_selectRONEnH$PACS)[,2] /  table(xdata_selectRONEnH$SevMax6)
barplot(t(Sev),main="",
        xlab="COVID-19 severity", col=c("#ffa44a","#c7264f"),legend = c("NO PACS", "PACS"),)
barplot(Sev_per,main="",ylab="PACS (%)",
        xlab="COVID-19 severity", col=c("#c7264f"), ylim=c(0,1))




#PACS score----

PACS_model <- glm(PACS ~ scale(Age) + Nr_Symptoms + Asthma + IgM * IgG3, 
                  xdata_selectRONEnH, family=binomial(link="logit"), x=TRUE)

PACS_predictions <- predict(PACS_model, xdata_selectRONEnH, type="response")
per_roc_LC1_Asthma_IgG3_IgM_IgA1_Sym <-roc(PACS ~ PACS_predictions, xdata_selectRONEnH, ci=TRUE) 


ggcoefstats(PACS_model, output="plot")
ggcoefstats(PACS_model, output="tidy")

#__Primary infection----
PACS_model_sh <- shrink(PACS_model, type = "global", method = "dfbeta")

PACS_predictions_outpatients <- predict(PACS_model_sh, xdata_selectRONEnHosp, type="response")
roc_PACS_outpatients <-roc(PACS ~ PACS_predictions_outpatients, xdata_selectRONEnHosp, ci=TRUE) 

PACS_predictions_hospitalized <- predict(PACS_model_sh, xdata_selectRONEHosp, type="response")
roc_PACS_hospitalized <-roc(PACS ~ PACS_predictions_hospitalized, xdata_selectRONEHosp, ci=TRUE) 


pROC::plot.roc(per_roc_LC1_Asthma_IgG3_IgM_IgA1_Sym, print.auc=TRUE, ci=TRUE, legacy.axes=TRUE, print.auc.y = .6,print.auc.x = .4)
pROC::plot.roc(roc_PACS_outpatients, print.auc=TRUE, ci=TRUE, legacy.axes=TRUE, axes=TRUE, print.auc.x = .4, add=F,print.auc.col=c("#4D77EB"), col=c("#4D77EB"))
pROC::plot.roc(roc_PACS_hospitalized, print.auc=TRUE, ci=TRUE, legacy.axes=TRUE, print.auc.y = .4,print.auc.x = .4, add=TRUE,print.auc.col=c("#E64A4A"),col=c("#E64A4A") )


calib1 <- val.prob.ci.2(PACS_predictions, as.numeric(xdata_selectRONEnH$PACS) - 1, 
                        pl = TRUE)

calibrate.plot(y = xdata_selectRONEnH$PACS, p = PACS_predictions, 
               las=1, shade.col= "lightgrey",  rug.par = list(side = 1, col = "black", lwd= 2))

Brier1 <-calib1[11]
Int1 <-calib1[12]
Slope1 <-calib1[13]
grid.text(paste("Brier score =", round(as.numeric(Brier1), 3)), just = "left", x=0.15, y=unit(0.7, "npc"), gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("Calibration-in-the-large =", round(as.numeric(Int1), 2)), just = "left", x=0.15, y=unit(0.8, "npc"), gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("Calibration slope =", round(as.numeric(Slope1), 2)), just = "left", x=0.15, y=unit(0.75, "npc"), gp=gpar(fontsize=11.5, col="black"))



#__6-month follow-up----
PACS_predictions_6m <- predict(PACS_model_sh, xdata_selectRSIXnH, type="response")
roc_PACS_6m <-roc(PACS ~ PACS_predictions_6m, xdata_selectRSIXnH, ci=TRUE) 

pROC::plot.roc(roc_PACS_6m, print.auc=TRUE, ci=TRUE, legacy.axes=TRUE,  print.auc.y = .6,print.auc.x = .4)

calib2 <- val.prob.ci.2(PACS_predictions_6m, as.numeric(xdata_selectRSIXnH$PACS) - 1, 
                        pl = TRUE)

calibrate.plot(y = xdata_selectRSIXnH$PACS, p = PACS_predictions_6m, 
               las=1, shade.col= "lightgrey",  rug.par = list(side = 1, col = "black", lwd= 2))

Brier2 <-calib2[11]
Int2 <-calib2[12]
Slope2 <-calib2[13]
grid.text(paste("Brier score =", round(as.numeric(Brier2), 3)), just = "left", x=0.15, y=unit(0.7, "npc"), gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("Calibration-in-the-large =", round(as.numeric(Int2), 2)), just = "left", x=0.15, y=unit(0.8, "npc"), gp=gpar(fontsize=11.5, col="black"))
grid.text(paste("Calibration slope =", round(as.numeric(Slope2), 2)), just = "left", x=0.15, y=unit(0.75, "npc"), gp=gpar(fontsize=11.5, col="black"))


#__Decision curve analysis----

xdata_selectRONEnH_2 <-
  xdata_selectRONEnH %>%
  mutate(prob_PACS= predict(PACS_model, xdata_selectRONEnH, type="response"))

dca(PACS ~ 
      prob_PACS , 
    xdata_selectRONEnH_2) %>%
  plot(smooth = TRUE)

dca(PACS ~ 
      prob_PACS , 
    xdata_selectRONEnH_2 %>% filter(Hospitalized == "YES" )) %>%
  plot(smooth = TRUE)

#__Interaction plots----
interact_plot(PACS_model, pred = IgG3, modx = IgM, modx.values = "plus-minus", interval = TRUE, int.width = 0.8)    
interact_plot(PACS_model, pred = IgM, modx = IgG3, modx.values = "plus-minus", interval = TRUE, int.width = 0.8)





#PACS score code for implementation----

xdata_selectRONEnHnew = xdata_selectRONEnH %>% 
  mutate( Asthma = ifelse(Asthma == "YES", 1, 0))

PACS_score <- glm(PACS ~ -1 + offset(-0.981011 + 0.2616998 *scale(Age) + 
                                                  0.3307986 * Nr_Symptoms + 
                                                  1.896502 * Asthma + 
                                                  0.8429766 * IgM  + 
                                                  1.3716198 * IgG3 +
                                                  -1.5316550 * IgM * IgG3), family = binomial, data = xdata_selectRONEnHnew)

new_PACS_score_prediction <- predict(PACS_score, xdata_selectRONEnHnew, type="response")

roc_PACS_score <-roc(PACS ~ new_PACS_score_prediction, xdata_selectRONEnHnew, ci=TRUE) 

pROC::plot.roc(roc_PACS_score, print.auc=TRUE, ci=TRUE, legacy.axes=TRUE)