In dmontemayor/Rcricvol: Data Processing Tools in R for cricvol project

Load data

#untar <- read.csv("../data/zscore_untargeted.csv")
clin <- read.csv("../data/partitioning.csv")
names(clin)[1] <- "patients"
#untarmet <- read.csv("../data/CRIC_clinical_creatinine_normalized_samplemetadata.csv", header = FALSE)

#jing <- read.csv("../data/60 var selected by 8 models from lasso RF TB (models with both ions and cli).csv")
#jing_ions <- as.numeric(unique(substring(jing$Ion, 5,15)))

# save copy of original untargeted ions 
#original_untar <- untar

# only include V3Y0
#untar <- untar[untar$visit == "V3Y0",]

#pats <- untar$patientid 

# include only jing ions
#untar <- untar[,names(untar) %in% paste0("V", jing_ions)]

# combine untargeted and clinical data
#untar$patients <- pats
#combined <- merge(untar, clin, by = "patients")


# separate into groups, remove patient information
#train1 <- combined[combined$group == 0,][-1]
#train2 <- combined[combined$group == 1,][-1]
#train3 <- combined[combined$group == 2,][-1]

# metadata <- c("group", "rawdata.mu_egfr", "rawdata.sigma_egfr", "rawdata.r_group", "rawdata.v_group")

#write.csv(train1, "../data/clin_and_untar_train1.csv", row.names = FALSE)
#write.csv(train2, "../data/clin_and_untar_train2.csv", row.names = FALSE)
#write.csv(train3, "../data/clin_and_untar_train3.csv", row.names = FALSE)

#clin1 <- clin[clin$group == 0,][-1]
#clin2 <- clin[clin$group == 1,][-1]
#clin3 <- clin[clin$group == 2,][-1]

#write.csv(clin1, "../data/clin_train1.csv", row.names = FALSE)
#write.csv(clin2, "../data/clin_train2.csv", row.names = FALSE)
#write.csv(clin3, "../data/clin_train3.csv", row.names = FALSE)

clin1 <- read.csv("../data/clin_train1.csv")
clin2 <- read.csv("../data/clin_train2.csv")
clin3 <- read.csv("../data/clin_train3.csv")

train1 <- read.csv("../data/clin_and_untar_train1.csv")
train2 <- read.csv("../data/clin_and_untar_train2.csv")
train3 <- read.csv("../data/clin_and_untar_train3.csv")

pca_coords <- read.csv("../data/pca_coords.csv")
pca_coords_t1 <- merge(clin[clin$group == 0,], pca_coords, by = "patients")[-1]
pca_coords_t2 <- merge(clin[clin$group == 1,], pca_coords, by = "patients")[-1]
pca_coords_t3 <- merge(clin[clin$group == 2,], pca_coords, by = "patients")[-1]

tsne_coords <- read.csv("../data/tsne_coords.csv")
tsne_coords_t1 <- merge(clin[clin$group == 0,], tsne_coords, by = "patients")[-1]
tsne_coords_t2 <- merge(clin[clin$group == 1,], tsne_coords, by = "patients")[-1]
tsne_coords_t3 <- merge(clin[clin$group == 2,], tsne_coords, by = "patients")[-1]

som_coords <- read.csv("../data/som_coords.csv")
som_coords_t1 <- merge(clin[clin$group == 0,], som_coords, by = "patients")[-1]
som_coords_t2 <- merge(clin[clin$group == 1,], som_coords, by = "patients")[-1]
som_coords_t3 <- merge(clin[clin$group == 2,], som_coords, by = "patients")[-1]

Linear Models: Clinical and Ions

#  EGFR Rate with Clinical Variables
lcr <- lm_rate_3xCV(clin1, clin2, clin3)

# Predicting EGFR Volatility with Clinical Variables
lcv <- lm_vol_3xCV(clin1, clin2, clin3)

# Predicting EGFR Rate with Clinical Variables and Untargeted Ions
lbr <- lm_rate_3xCV(train1, train2, train3)

# Predicting EGFR Volatility with combical Variables and Untargeted Ions
lbv <- lm_vol_3xCV(train1, train2, train3)

Random Forest: Clinical and Ions

#  EGFR Rate with Clinical Variables
rfcr <- rf_rate_3xCV(clin1, clin2, clin3)

# Predicting EGFR Volatility with Clinical Variables
rfcv <- rf_vol_3xCV(clin1, clin2, clin3)

# Predicting EGFR Rate with Clinical Variables and Untargeted Ions
rfbr <- rf_rate_3xCV(train1, train2, train3)

# Predicting EGFR Volatility with combical Variables and Untargeted Ions
rfbv <- rf_vol_3xCV(train1, train2, train3)

Plot Clin vs Clin and Ions

plot_compare_models(linear_clin = lcr$res1, linear_other = lbr$res1, rf_clin = rfcr$res1, rf_other = rfbr$res1, description_other = "Ions", predicting = "EGFR Rate")
plot_compare_models(linear_clin = lcv$res1, linear_other = lbv$res1, rf_clin = rfcv$res1, rf_other = rfbv$res1, description_other = "Ions", predicting = "EGFR Volatility")

Clin vs Clin and PCA Coordinates

# Linear: rate
pca_lr <- lm_rate_3xCV(pca_coords_t1, pca_coords_t2, pca_coords_t3)

# Linear: vol
pca_lv <- lm_vol_3xCV(pca_coords_t1, pca_coords_t2, pca_coords_t3)

# RF: rate
pca_rfr <- rf_rate_3xCV(pca_coords_t1, pca_coords_t2, pca_coords_t3)

# RF: vol
pca_rfv <- rf_vol_3xCV(pca_coords_t1, pca_coords_t2, pca_coords_t3)

# plot
plot_compare_models(linear_clin = lcr$res1, linear_other = pca_lr$res1, rf_clin = rfcr$res1, rf_other = pca_rfr$res1, description_other = "Ions", predicting = "EGFR Rate")
plot_compare_models(linear_clin = lcv$res1, linear_other = pca_lv$res1, rf_clin = rfcv$res1, rf_other = pca_lv$res1, description_other = "Ions", predicting = "EGFR Volatility")

TSNE

# Linear: rate
tsne_lr <- lm_rate_3xCV(tsne_coords_t1, tsne_coords_t2, tsne_coords_t3)

# Linear: vol
tsne_lv <- lm_vol_3xCV(tsne_coords_t1, tsne_coords_t2, tsne_coords_t3)

# RF: rate
tsne_rfr <- rf_rate_3xCV(tsne_coords_t1, tsne_coords_t2, tsne_coords_t3)

# RF: vol
tsne_rfv <- rf_vol_3xCV(tsne_coords_t1, tsne_coords_t2, tsne_coords_t3)

# plot
plot_compare_models(linear_clin = lcr$res1, linear_other = tsne_lr$res1, rf_clin = rfcr$res1, rf_other = tsne_rfr$res1, description_other = "Ions", predicting = "EGFR Rate")
plot_compare_models(linear_clin = lcv$res1, linear_other = tsne_lv$res1, rf_clin = rfcv$res1, rf_other = tsne_lv$res1, description_other = "Ions", predicting = "EGFR Volatility")

SOM

# Linear: rate
som_lr <- lm_rate_3xCV(som_coords_t1, som_coords_t2, som_coords_t3)

# Linear: vol
som_lv <- lm_vol_3xCV(som_coords_t1, som_coords_t2, som_coords_t3)

# RF: rate
som_rfr <- rf_rate_3xCV(som_coords_t1, som_coords_t2, som_coords_t3)

# RF: vol
som_rfv <- rf_vol_3xCV(som_coords_t1, som_coords_t2, som_coords_t3)

# plot
plot_compare_models(linear_clin = lcr$res1, linear_other = som_lr$res1, rf_clin = rfcr$res1, rf_other = som_rfr$res1, description_other = "Ions", predicting = "EGFR Rate")
plot_compare_models(linear_clin = lcv$res1, linear_other = som_lv$res1, rf_clin = rfcv$res1, rf_other = som_lv$res1, description_other = "Ions", predicting = "EGFR Volatility")