Quantitative pathology and APOE genotype reveal dementia risk and progression in Lewy body disease.

Code ↔ Paper

7 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 7 matches

[1] § Materials and methods › Logistic regression modelling of dementia risk ↔ QuantPath_Figure_4_and_5.R, lines 90 to 130 · score 0.87 · logistic regression models, continuous predictors, predicted probabilities, categorical variables, Marginal, binomial · Report a mismatch
[2] § Materials and methods › Pathology burden classification and diagnostic performance metrics ↔ QuantPath_Figure_1_and_2.R, lines 1212 to 1270 · score 0.78 · pathology presence ratio, pathology absence ratio, odds ratio, PPR, PAR, PPV · Report a mismatch
[3] § Materials and methods › Statistical analysis ↔ SuStaIn_ASyn_Subtype.ipynb, lines 446 to 508 · score 0.75 · AMY_LS, AMY_MI, DA, MFG, MTG, PTEGM · Report a mismatch
[4] § Materials and methods › Neuropathological assessment ↔ QuantPath_Table1_and_Suppl_Tables_3_4_5.R, lines 553 to 595 · score 0.73 · NIA AA, Thal phase, ischaemic pathology, Braak stage, ADNC, pTau · Report a mismatch
[5] § Materials and methods › Neuropathological assessment ↔ QuantPath_Figure_1_and_2.R, lines 1212 to 1270 · score 0.67 · pathology absence ratio, odds ratio, PPR, PAR, Segments, PPV · Report a mismatch
[6] § Materials and methods › Statistical analysis ↔ QuantPath_Figure_4_and_5.R, lines 90 to 130 · score 0.66 · logistic regression models, predicted probabilities, biological sex, death, duration, age · Report a mismatch
[7] § Materials and methods › Statistical analysis ↔ QuantPath_Table1_and_Suppl_Tables_3_4_5.R, lines 553 to 595 · score 0.62 · Kruskal Wallis, Thal phase, Lewy bodies, Braak stage, density, ADNC · Report a mismatch

Paper

Loaded from Europe PMC by your browser; OSCR does not store it: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

R · 728 lines · 23 KB · CC-BY-4.0 · 2 matches

#MULTIVARIATE LOGISTICAL REGRESSUION#####################
# Load required libraries
library(readxl)
library(tidyverse)
library(ggplot2)
library(gridExtra)
library(reshape2)
library(FactoMineR)
library(factoextra)
library(dplyr)
library(ggeffects)
library(gtsummary)
library(flextable)
# Set directories
setwd("DIR")
datadir <- "DIR"
dir.create(datadir, showWarnings = FALSE)
setwd(datadir)
# Load data
ex <- read_excel("EXCEL",
sheet = "SHEET")
excel <- as.data.frame(ex)
excel_main <- excel
########### Bin variables: Duration>15 and Age_at_Death>77
# Load dplyr package
library(dplyr)
# Filter the dataframe to remove rows where ApoE and INFARCT have NA values
# Ensure no missing values
excel_main <- excel_main %>% filter(!is.na(Duration))
# Calculate the median of Duration
median_duration <- median(excel_main$Duration, na.rm = TRUE, 0)
print(median_duration)
# Bin the values at the median
excel_main$Duration_Binned <- cut(excel_main$Duration,
breaks = c(-Inf, median_duration, Inf),
labels = c(paste0("< ", median_duration, "yrs"),
paste0("≥ ", median_duration, "yrs")), # Use median in labels
include.lowest = TRUE, right = FALSE) # left-inclusive binning
table(excel_main$Duration_Binned)
# Ensure no missing values
excel_main <- excel_main %>% filter(!is.na(Age_at_Death))
# Calculate the median of Age_at_Death
median_Age_at_Death <- median(excel_main$Age_at_Death, na.rm = TRUE)
print(median_Age_at_Death)
# Bin the values at the median
excel_main$Age_at_Death_Binned <- cut(excel_main$Age_at_Death,
breaks = c(-Inf, median_Age_at_Death, Inf),
labels = c(paste0("< ", median_Age_at_Death, "yrs"),
paste0("≥ ", median_Age_at_Death, "yrs")), # Use median in labels
include.lowest = TRUE, right = FALSE) # left-inclusive binning
table(excel_main$Age_at_Death_Binned)
# Convert categorical variables to factors
excel_main$Duration_Binned <- as.factor(excel_main$Duration_Binned)
excel_main$Age_at_Death_Binned <- as.factor(excel_main$Age_at_Death_Binned)
excel_main$INFARCT <- as.factor(excel_main$INFARCT)
excel_main$ApoE <- as.factor(excel_main$ApoE)
# View first few rows to check binning
head(excel_main[, c("Duration", "Duration_Binned", "Age_at_Death", "Age_at_Death_Binned")])
# View first few rows to check binning
head(excel_main)
####################################################FULL_MODEL
UPPR_DIR <- paste0(datadir,"/FULL_MODEL")
dir.create(UPPR_DIR)
setwd(UPPR_DIR)
excel_main <- subset(excel_main, select = c(DEMENTIA, ApoE, ASYN_FRNT_HIPP, AT8_FRNT_HIPP, ABETA_FRNT_HIPP, INFARCT, Duration_Binned, Age_at_Death_Binned, BIOLOGICAL_SEX))
# Define predictors and outcome
predictors <- c("ASYN_FRNT_HIPP", "AT8_FRNT_HIPP", "ABETA_FRNT_HIPP", "INFARCT", "Duration_Binned", "Age_at_Death_Binned", "BIOLOGICAL_SEX")
outcome <- "DEMENTIA"
# Ensure outcome variable is a factor
excel_main[[outcome]] <- as.factor(excel_main[[outcome]])
####################################################APOE
# Manually define x-axis limits for continuous predictors
xlims_list <- list(
"ASYN_FRNT_HIPP" = c(0, 50),
"ABETA_FRNT_HIPP" = c(0, 8),
"AT8_FRNT_HIPP" = c(0, 15)
)
# Run the analysis separately for ApoE3 and ApoE4
for (apoe_type in c("ApoE3", "ApoE4")) {
# Filter dataset for current ApoE type
#apoe_type <- "ApoE3"
subset_data <- excel_main %>% filter(ApoE == apoe_type)
# Train logistic regression model
model_base <- glm(DEMENTIA ~ ., data = subset_data[, c(predictors, outcome)], family = "binomial")
library(car)
vif(model_base)
library(sjPlot)
library(sjmisc)
library(ggeffects)
library(cowplot)
library(RColorBrewer)
library(ggplot2)
# Generate marginal effects for all predictors
effects_list <- ggpredict(model_base) # Get predicted probabilities
plot_order <- c(
"ASYN_FRNT_HIPP", "ABETA_FRNT_HIPP", "AT8_FRNT_HIPP",
"BIOLOGICAL_SEX", "INFARCT", "Duration_Binned", "Age_at_Death_Binned"
)
effects_list <- effects_list[plot_order]
# Define categorical variables
categorical_vars <- c("INFARCT", "Age_at_Death_Binned", "Duration_Binned", "BIOLOGICAL_SEX")
# Define colors for continuous variable lines
continuous_colors <- c("α-synuclein" = "forestgreen",
"Aβ" = "firebrick",
"pTau" = "dodgerblue")
# Define x-axis label mapping
x_axis_labels <- c(
"ASYN_FRNT_HIPP" = "α-synuclein",
"ABETA_FRNT_HIPP" = "Aβ",
"AT8_FRNT_HIPP" = "pTau",
"INFARCT" = "Ischaemia",
"BIOLOGICAL_SEX" = "Biological sex",
"Duration_Binned" = "Duration",
"Age_at_Death_Binned" = "Age at death"
)
# Recode levels for INFARCT
infarct_labels <- c("INF_0" = "Absent", "INF_1" = "Present")
bsex_labels <- c("M" = "Male", "F" = "Female")
# Create individual plots
common_ylim <- c(0, 100)
plot_list <- lapply(seq_along(effects_list), function(i) {
var_name <- names(effects_list)[i]
effects_data <- as.data.frame(effects_list[[i]])
if (nrow(effects_data) == 0) return(NULL) # Skip empty frames
is_categorical <- var_name %in% categorical_vars
if (!is_categorical) {
plot_obj <- ggplot(effects_data, aes(x = x, y = predicted * 100)) +
geom_ribbon(
aes(ymin = conf.low * 100, ymax = conf.high * 100),
fill = continuous_colors[[x_axis_labels[[var_name]]]],
alpha = 0.2
) +
geom_line(
size = 3,
color = continuous_colors[[x_axis_labels[[var_name]]]]
) +
labs(
x = x_axis_labels[[var_name]],
y = if (i == 1) "Dementia probability (%)" else NULL
) +
xlim(xlims_list[[var_name]]) + # 👈 Apply manual x-axis limit here
ylim(common_ylim) +
theme_minimal() +
theme(
plot.title = element_blank(),
axis.title.x = element_text(size = 34),
axis.title.y = element_text(size = 34),
axis.text = element_text(size = 34),
plot.margin = margin(t = 40, r = 5, b = 5, l = 5, unit = "pt"),
legend.position = "none"
)
} else {
if (var_name == "INFARCT") {
effects_data$x <- factor(effects_data$x,
levels = names(infarct_labels),
labels = infarct_labels)
} else if (var_name == "BIOLOGICAL_SEX") {
effects_data$x <- factor(effects_data$x,
levels = names(bsex_labels),
labels = bsex_labels)
}
plot_obj <- ggplot(effects_data, aes(x = x, y = predicted * 100, color = as.factor(x))) +
geom_point(size = 10) +
geom_errorbar(aes(ymin = conf.low * 100, ymax = conf.high * 100),
width = 0.7, linetype = "dotted", size = 2) +
scale_color_manual(values = c("orchid1", "darkorchid4")) +
labs(
x = x_axis_labels[[var_name]],
y = if (i == 1) "Dementia probability (%)" else NULL,
color = "Category"
) +
ylim(common_ylim) +
theme_minimal() +
theme(
plot.title = element_blank(),
axis.title.x = element_text(size = 34),
axis.title.y = element_text(size = 34),
axis.text = element_text(size = 34),
plot.margin = margin(t = 40, r = 5, b = 5, l = 5, unit = "pt"),
legend.position = "none"
)
}
# Remove y-axis title for plots after the first one
if (i != 1) {
plot_obj <- plot_obj + theme(axis.title.y = element_blank())
}
return(plot_obj)
})
# Remove NULLs from empty predictors
plot_list <- plot_list[!sapply(plot_list, is.null)]
# Arrange plots
sjplot <- cowplot::plot_grid(plotlist = plot_list, nrow = 1, align = "h")
# Display plot
print(sjplot)
# Save
ggsave(paste0("sjplot_UPPERQUARTILE_", apoe_type, ".png"), sjplot, width = 35, height = 5, dpi = 300)
}
#######################DECISION TREE################
# Define continuous variables
continuous_vars <- c("ASYN_FRNT_HIPP", "ABETA_FRNT_HIPP", "AT8_FRNT_HIPP")
# Apply per group
get_fixed_splits_by_apoe <- function(data, apoe_group) {
df_sub <- data %>%
filter(ApoE == apoe_group) %>%
mutate(DEMENTIA_BIN = ifelse(DEMENTIA == "Dem", 1, 0))
# Define cutoffs
asyn_75 <- quantile(df_sub$ASYN_FRNT_HIPP, 0.75, na.rm = TRUE)
abeta_75 <- quantile(df_sub$ABETA_FRNT_HIPP, 0.75, na.rm = TRUE)
at8_85 <- quantile(df_sub$AT8_FRNT_HIPP, 0.85, na.rm = TRUE)
tibble(
ApoE = apoe_group,
variable = c("ASYN_FRNT_HIPP", "ABETA_FRNT_HIPP", "AT8_FRNT_HIPP"),
percentile = c("75th", "75th", "85th"),
fixed_cut = c(asyn_75, abeta_75, at8_85)
)
}
# Run for both ApoE groups
split_apoe3 <- get_fixed_splits_by_apoe(df_binned, "APOE ε3")
split_apoe4 <- get_fixed_splits_by_apoe(df_binned, "APOE ε4")
# Combine
split_results_all <- bind_rows(split_apoe3, split_apoe4)
# View results
print(split_results_all)
# Save to CSV
write_csv(split_results_all, "logistic_best_splits_by_apoe.csv")
# Define the fixed splits for each ApoE group
fixed_splits <- split_results_all %>%
arrange(ApoE, variable) %>%
select(ApoE, variable, fixed_cut) %>%
mutate(fixed_cut = round(fixed_cut, 2))
df_binned <- df_binned %>%
mutate(DEMENTIA_BIN = ifelse(DEMENTIA == "Dem", 1, 0))
###DT_apoe3
# -- 1. Rename variables in importance table
result_apoe3_renamed <- as.data.frame(result_apoe3$importance) %>%
mutate(Variable = case_when(
Variable == "INFARCTINF_1" ~ "INFARCT",
Variable == "BIOLOGICAL_SEXM" ~ "BIOLOGICAL_SEX",
Variable == "Age_at_Death_Binned≥ 77" ~ "Age_at_Death_Binned",
Variable == "Duration_Binned≥ 15" ~ "Duration_Binned",
TRUE ~ Variable
))
# -- 2. Get ordered variables by importance
importance_apoe3 <- result_apoe3_renamed %>%
arrange(desc(Importance)) %>%
pull(Variable)
# -- 3. Merge with fixed splits
ordered_variables_apoe3 <- fixed_splits %>%
filter(ApoE == "APOE ε3") %>%
arrange(match(variable, importance_apoe3))
ordered_vars <- importance_apoe3
var_rename <- c(
ASYN_FRNT_HIPP = "α-synuclein",
ABETA_FRNT_HIPP = "Aβ",
Duration_Binned = "Duration",
INFARCT = "Ischaemia",
Age_at_Death_Binned = "Age at death",
BIOLOGICAL_SEX = "Biological sex",
AT8_FRNT_HIPP = "pTau"
)
# Apply renaming
ordered_vars <- unname(var_rename[ordered_vars])
ordered_vars
# -- 4. Create binned columns
cutoffs <- ordered_variables_apoe3$fixed_cut
names(cutoffs) <- ordered_variables_apoe3$variable
df_binned1 <- df_binned %>%
mutate(
ASYN_FRNT_HIPP_CAT = ifelse(ASYN_FRNT_HIPP >= cutoffs["ASYN_FRNT_HIPP"], paste0("\u2265", cutoffs["ASYN_FRNT_HIPP"]), paste0("<", cutoffs["ASYN_FRNT_HIPP"])),
ABETA_FRNT_HIPP_CAT = ifelse(ABETA_FRNT_HIPP >= cutoffs["ABETA_FRNT_HIPP"], paste0("\u2265", cutoffs["ABETA_FRNT_HIPP"]), paste0("<", cutoffs["ABETA_FRNT_HIPP"])),
AT8_FRNT_HIPP_CAT = ifelse(AT8_FRNT_HIPP >= cutoffs["AT8_FRNT_HIPP"], paste0("\u2265", cutoffs["AT8_FRNT_HIPP"]), paste0("<", cutoffs["AT8_FRNT_HIPP"]))
)
# -- 5. Prepare dataset for APOE ε3 group
df_apoe3 <- df_binned1 %>%
filter(ApoE == "APOE ε3") %>%
mutate(
`α-synuclein` = ASYN_FRNT_HIPP_CAT,
`Aβ` = ABETA_FRNT_HIPP_CAT,
`pTau` = AT8_FRNT_HIPP_CAT,
Ischaemia = factor(INFARCT, levels = c("INF_0", "INF_1"), labels = c("Absent", "Present")),
`Biological sex` = factor(BIOLOGICAL_SEX, levels = c("F", "M"), labels = c("FEMALE", "MALE")),
`Age at death` = Age_at_Death_Binned,
Duration = Duration_Binned
) %>%
select(DEMENTIA, DEMENTIA_BIN, `α-synuclein`, `Aβ`, `pTau`, Ischaemia, `Biological sex`, `Age at death`, Duration) %>%
mutate(across(everything(), as.character))
# -- 6. Evaluate path-level risk
library(dplyr)
# -- 1. Evaluate path-level risk
tree_paths_eval_apoe3 <- df_apoe3 %>%
group_by(`α-synuclein`, `Aβ`, `pTau`, Ischaemia, `Biological sex`, `Age at death`, Duration) %>%
summarise(
n = n(),
pct_high = mean(DEMENTIA_BIN == 1, na.rm = TRUE),
.groups = "drop"
) %>%
mutate(
risk_label = case_when(
is.na(pct_high) ~ "No data",
pct_high >= 0.7 ~ "High risk",
TRUE ~ "Low risk"
),
label = paste0(risk_label, "\n(n=", n, ", ",
ifelse(is.na(pct_high), "NA", round(pct_high * 100)), "%)"),
display_label = risk_label
)
write.csv(tree_paths_eval_apoe3, "ApoE3_risk_bins_OBS.csv", row.names = FALSE)
# -- 2. Define ordered variables for tree depth (same as image structure)
ordered_vars
# -- 3. Create custom PathString that stops branching when pct_high ≥ 0.7 <20
tree_paths_eval_apoe3$PathString <- apply(tree_paths_eval_apoe3, 1, function(row) {
path <- c("Start")
label_to_use <- row[["display_label"]] # default
for (v in ordered_vars) {
value <- row[[v]]
path <- c(path, paste0(v, ": ", value))
# Check if this subgroup exceeds threshold
subset_rows <- tree_paths_eval_apoe3
for (i in seq_along(path)[-1]) {
parts <- strsplit(path[i], ": ")[[1]]
subset_rows <- subset_rows[subset_rows[[parts[1]]] == parts[2], ]
}
if (nrow(subset_rows) > 0) {
avg_pct <- mean(subset_rows$pct_high)
if (avg_pct >= 0.6) {
label_to_use <- "High risk"
break
} else if (avg_pct <= 0.30) {
label_to_use <- "Low risk"
break
}
}
} # <-- closes the for loop
paste(c(path, label_to_use), collapse = "/")
})
library(dplyr)
library(DiagrammeR)
library(stringr)
library(digest)
# ---- Step 1: Extract node paths from tree_paths_eval
paths <- strsplit(tree_paths_eval_apoe3$PathString, "/")
# ---- Step 2: Create edges and unique node labels
edges <- data.frame(from = character(), to = character(), label = character())
nodes <- data.frame(id = character(), label = character(), group = character(), stringsAsFactors = FALSE)
for (p in paths) {
for (i in 2:length(p)) {
from <- paste(p[1:(i - 1)], collapse = "/")
to <- paste(p[1:i], collapse = "/")
label <- p[i]
# Generate unique, clean IDs for Graphviz
from_id <- paste0("node_", digest(from, algo = "crc32"))
to_id <- paste0("node_", digest(to, algo = "crc32"))
edges <- rbind(edges, data.frame(from = from_id, to = to_id, label = label))
group <- ifelse(grepl("risk", label), "risk",
ifelse(grepl("^α-syn", label), "asyn",
ifelse(grepl("^Aβ", label), "abeta",
ifelse(grepl("^pTau", label), "ptau",
ifelse(grepl("Ischa", label), "isch",
ifelse(grepl("Sex", label), "sex",
ifelse(grepl("Duration", label), "dur", "default")))))))
nodes <- rbind(nodes, data.frame(
id = to_id,
label = label,
group = group,
stringsAsFactors = FALSE
))
}
}
# Remove duplicates
nodes <- nodes[!duplicated(nodes$id), ]
edges <- edges[!duplicated(edges), ]
# Get root node ID from the first edge
root_id <- edges$from[1]
# If it's not in the node list, add it manually
if (!(root_id %in% nodes$id)) {
nodes <- rbind(nodes, data.frame(
id = root_id,
label = "APOE ε3",
group = "apoe",
stringsAsFactors = FALSE
))
}
# ---- Step 3: Build Graphviz DOT syntax
color_map <- c(
"asyn" = "lawngreen",
"abeta" = "lightpink",
"isch" = "khaki1",
"ptau" = "lightskyblue1",
"sex" = "plum1",
"dur" = "lemonchiffon",
"risk" = "mistyrose",
"apoe" = "dodgerblue1", # ← custom color for APOE root
"default" = "white"
)
# Fix labels: escape quotes, remove newlines for Graphviz compatibility
nodes$label <- gsub("\"", "'", nodes$label) # Replace double quotes
nodes$label <- gsub("\n", "\\n", nodes$label) # Escape line breaks
node_defs <- paste0(
nodes$id, ' [label="', nodes$label,
'", style=filled, fillcolor="', color_map[nodes$group], '"];'
)
edge_defs <- paste0(
edges$from, ' -> ', edges$to, ';'
)
# ---- Step 4: Wrap in grViz
dot_code_apoe3 <- paste0(
"digraph G {\n",
" graph [layout = dot, rankdir = TB];\n",
" node [shape=box, fontname=Helvetica];\n\n",
paste(node_defs, collapse = "\n"), "\n\n",
paste(edge_defs, collapse = "\n"), "\n}"
)
# ---- Step 5: Render the final publication-style graph
grViz(dot_code_apoe3)
library(rsvg)
library(DiagrammeRsvg)
# Render and export to PDF
grViz(dot_code_apoe3) %>%
export_svg() %>%
charToRaw() %>%
rsvg_pdf("decision_tree_apoe3_min_split_ObsOnly.pdf")
###DT_apoe4
# -- 1. Rename variables in importance table
result_apoe4_renamed <- as.data.frame(result_apoe4$importance) %>%
mutate(Variable = case_when(
Variable == "INFARCTINF_1" ~ "INFARCT",
Variable == "BIOLOGICAL_SEXM" ~ "BIOLOGICAL_SEX",
Variable == "Age_at_Death_Binned≥ 77" ~ "Age_at_Death_Binned",
Variable == "Duration_Binned≥ 15" ~ "Duration_Binned",
TRUE ~ Variable
))
# -- 2. Get ordered variables by importance
importance_apoe4 <- result_apoe4_renamed %>%
arrange(desc(Importance)) %>%
pull(Variable)
# -- 3. Merge with fixed splits
ordered_variables_apoe4 <- fixed_splits %>%
filter(ApoE == "APOE ε4") %>%
arrange(match(variable, importance_apoe4))
ordered_vars <- importance_apoe4
var_rename <- c(
ASYN_FRNT_HIPP = "α-synuclein",
ABETA_FRNT_HIPP = "Aβ",
Duration_Binned = "Duration",
INFARCT = "Ischaemia",
Age_at_Death_Binned = "Age at death",
BIOLOGICAL_SEX = "Biological sex",
AT8_FRNT_HIPP = "pTau"
)
# Apply renaming
ordered_vars <- unname(var_rename[ordered_vars])
# -- 4. Create binned columns
cutoffs <- ordered_variables_apoe4$fixed_cut
names(cutoffs) <- ordered_variables_apoe4$variable
df_binned1 <- df_binned %>%
mutate(
ASYN_FRNT_HIPP_CAT = ifelse(ASYN_FRNT_HIPP >= cutoffs["ASYN_FRNT_HIPP"], paste0("\u2265", cutoffs["ASYN_FRNT_HIPP"]), paste0("<", cutoffs["ASYN_FRNT_HIPP"])),
ABETA_FRNT_HIPP_CAT = ifelse(ABETA_FRNT_HIPP >= cutoffs["ABETA_FRNT_HIPP"], paste0("\u2265", cutoffs["ABETA_FRNT_HIPP"]), paste0("<", cutoffs["ABETA_FRNT_HIPP"])),
AT8_FRNT_HIPP_CAT = ifelse(AT8_FRNT_HIPP >= cutoffs["AT8_FRNT_HIPP"], paste0("\u2265", cutoffs["AT8_FRNT_HIPP"]), paste0("<", cutoffs["AT8_FRNT_HIPP"]))
)
# -- 5. Prepare dataset for APOE ε4 group
df_apoe4 <- df_binned1 %>%
filter(ApoE == "APOE ε4") %>%
mutate(
`α-synuclein` = ASYN_FRNT_HIPP_CAT,
`Aβ` = ABETA_FRNT_HIPP_CAT,
`pTau` = AT8_FRNT_HIPP_CAT,
Ischaemia = factor(INFARCT, levels = c("INF_0", "INF_1"), labels = c("Absent", "Present")),
`Biological sex` = factor(BIOLOGICAL_SEX, levels = c("F", "M"), labels = c("FEMALE", "MALE")),
`Age at death` = Age_at_Death_Binned,
Duration = Duration_Binned
) %>%
select(DEMENTIA, DEMENTIA_BIN, `α-synuclein`, `Aβ`, `pTau`, Ischaemia, `Biological sex`, `Age at death`, Duration) %>%
mutate(across(everything(), as.character))
# -- 6. Evaluate path-level risk
library(dplyr)
# -- 1. Evaluate path-level risk
tree_paths_eval_apoe4 <- df_apoe4 %>%
group_by(`α-synuclein`, `Aβ`, `pTau`, Ischaemia, `Biological sex`, `Age at death`, Duration) %>%
summarise(
n = n(),
pct_high = mean(DEMENTIA_BIN == 1, na.rm = TRUE),
.groups = "drop"
) %>%
mutate(
risk_label = case_when(
is.na(pct_high) ~ "No data",
pct_high >= 0.7 ~ "High risk",
TRUE ~ "Low risk"
),
label = paste0(risk_label, "\n(n=", n, ", ",
ifelse(is.na(pct_high), "NA", round(pct_high * 100)), "%)"),
display_label = risk_label
)
write.csv(tree_paths_eval_apoe4, "ApoE4_risk_bins_OBS.csv", row.names = FALSE)
# -- 2. Define ordered variables for tree depth (same as image structure)
ordered_vars
# -- 3. Create custom PathString that stops branching when pct_high ≥ 0.8
tree_paths_eval_apoe4$PathString <- apply(tree_paths_eval_apoe4, 1, function(row) {
path <- c("Start")
label_to_use <- row[["display_label"]] # default
for (v in ordered_vars) {
value <- row[[v]]
path <- c(path, paste0(v, ": ", value))
# Check if this subgroup exceeds threshold
subset_rows <- tree_paths_eval_apoe4
for (i in seq_along(path)[-1]) {
parts <- strsplit(path[i], ": ")[[1]]
subset_rows <- subset_rows[subset_rows[[parts[1]]] == parts[2], ]
}
if (nrow(subset_rows) > 0) {
avg_pct <- mean(subset_rows$pct_high)
if (avg_pct >= 0.66) {
label_to_use <- "High risk"
break
} else if (avg_pct <= 0.3) {
label_to_use <- "Low risk"
break
}
}
} # <-- closes the for loop
paste(c(path, label_to_use), collapse = "/")
})
library(dplyr)
library(DiagrammeR)
library(stringr)
library(digest)
# ---- Step 1: Extract node paths from tree_paths_eval
paths <- strsplit(tree_paths_eval_apoe4$PathString, "/")
# ---- Step 2: Create edges and unique node labels
edges <- data.frame(from = character(), to = character(), label = character())
nodes <- data.frame(id = character(), label = character(), group = character(), stringsAsFactors = FALSE)
for (p in paths) {
for (i in 2:length(p)) {
from <- paste(p[1:(i - 1)], collapse = "/")
to <- paste(p[1:i], collapse = "/")
label <- p[i]
# Generate unique, clean IDs for Graphviz
from_id <- paste0("node_", digest(from, algo = "crc32"))
to_id <- paste0("node_", digest(to, algo = "crc32"))
edges <- rbind(edges, data.frame(from = from_id, to = to_id, label = label))
group <- ifelse(grepl("risk", label), "risk",
ifelse(grepl("^α-syn", label), "asyn",
ifelse(grepl("^Aβ", label), "abeta",
ifelse(grepl("^pTau", label), "ptau",
ifelse(grepl("Ischa", label), "isch",
ifelse(grepl("Sex", label), "sex",
ifelse(grepl("Duration", label), "dur", "default")))))))
nodes <- rbind(nodes, data.frame(
id = to_id,
label = label,
group = group,
stringsAsFactors = FALSE
))
}
}
# Remove duplicates
nodes <- nodes[!duplicated(nodes$id), ]
edges <- edges[!duplicated(edges), ]
# Get root node ID from the first edge
root_id <- edges$from[1]
# If it's not in the node list, add it manually
if (!(root_id %in% nodes$id)) {
nodes <- rbind(nodes, data.frame(
id = root_id,
label = "APOE ε4",
group = "apoe",
stringsAsFactors = FALSE
))
}
# ---- Step 3: Build Graphviz DOT syntax
color_map <- c(
"asyn" = "lawngreen",
"abeta" = "lightpink",
"isch" = "khaki1",
"ptau" = "lightskyblue1",
"sex" = "plum1",
"dur" = "lemonchiffon",
"risk" = "mistyrose",
"apoe" = "firebrick2", # ← custom color for APOE root
"default" = "white"
)
node_defs <- paste0(
nodes$id, ' [label="', nodes$label,
'", style=filled, fillcolor="', color_map[nodes$group], '"];'
)
edge_defs <- paste0(
edges$from, ' -> ', edges$to, ';'
)
# ---- Step 4: Wrap in grViz
dot_code_apoe4 <- paste0(
"digraph G {\n",
" graph [layout = dot, rankdir = TB];\n",
" node [shape=box, fontname=Helvetica];\n\n",
paste(node_defs, collapse = "\n"), "\n\n",
paste(edge_defs, collapse = "\n"), "\n}"
)
# ---- Step 5: Render the final publication-style graph
grViz(dot_code_apoe4)
library(rsvg)
library(DiagrammeRsvg)
# Render and export to PDF
grViz(dot_code_apoe4) %>%
export_svg() %>%
charToRaw() %>%
rsvg_pdf("decision_tree_apoe4_min_split_ObsOnly.pdf")

QuantPath_Figure_4_and_5.R, under CC-BY-4.0 · at the source

Overview

Authors: Hemanth R Nelvagal^1,2,3, Nancy Chiraki^1,2,3, Toby Curless^3,4, Patrick W Cullinane^1,2,5,6, Alice Rockliffe^1,2, Samarth Pimparkar^1,2, Haruka Kawamura², Sophie Ollerenshaw², Isha Elahi², Sebastian Brandner^7,8, Lesley Wu¹, Raquel Real^1,3, Mina Ryten^3,9, John Hardy^3,7, Eduardo De Pablo Fernandez^1,2,5,10, Thomas T Warner^1,2,5,6, Huw R Morris^1,3,6, Yau Mun Lim^3,7, Zane Jaunmuktane^1,2,3

ORCID iDs: Hemanth R Nelvagal, Nancy Chiraki, Toby Curless, Patrick W Cullinane, Samarth Pimparkar, Haruka Kawamura, Isha Elahi, Sebastian Brandner, Lesley Wu, Raquel Real, Mina Ryten, John Hardy, Eduardo De Pablo Fernandez, Thomas T Warner, Huw R Morris, Yau Mun Lim, Zane Jaunmuktane

Institutions: National Hospital for Neurology and Neurosurgery (United Kingdom); Aligning Science Across Parkinson's (United States); UCL Queen Square Institute of Neurology (United Kingdom); University College London (United Kingdom); University College London Hospitals NHS Foundation Trust (United Kingdom); University of Cambridge (United Kingdom); UK Dementia Research Institute (United Kingdom); Queen Mary University of London (United Kingdom)

Journal: Brain : a journal of neurology, volume 149, issue 11, pages 3905-3922

Dates: received 12 September 2025; accepted 1 March 2026; published online 27 March 2026; in print November 2026

Type: Research article · Language: English

License: CC BY

Identifiers: DOI 10.1093/brain/awag114 · PMID 41889331 · PMCID PMC13730591 · OpenAlex W7141045465

Open access: hybrid, a free copy (OpenAlex)

Preprint: doi:10.1101/2025.10.20.25338378

Status: code verified (how the code is found and verified)

Categories: genetics / omics (modality), human (organism), Alzheimer's / dementia (population), Parkinson's (population)

Methods: Statistics, Machine learning, Preprocessing, Connectivity, fMRI & imaging

Keywords: Lewy body diseases, Parkinson’s disease, APOE genotype, orthostatic hypotension with ischaemic pathology, Alzheimer’s co-pathology, SuStaIn

MeSH: Apolipoproteins E*, Brain*, Dementia*, Lewy Body Disease*, Aged, Aged, 80 and over, alpha-Synuclein, Amyloid beta-Peptides, Apolipoprotein E4, Disease Progression, Female, Genotype, Humans, Male, Parkinson Disease, tau Proteins (* major topic)

Topic: Parkinson's Disease Mechanisms and Treatments (Clinical Neurology, Medicine), according to OpenAlex

Funding: Aligning Science Across Parkinson's (ASAP 000478); Michael J. Fox Foundation for Parkinson's Research

Citations: cited by 1 paper (Europe PMC); 64 references in the paper

Research resources: RRID:SCR_025020

Abstract

Dementia in Lewy body diseases (LBD) is common and arises through heterogeneous and incompletely understood pathways. Evidence suggests contributions from genetic factors, including APOE ɛ4 genotype, and co-pathology, including concomitant Alzheimer’s disease pathology and hypoperfusion related to orthostatic hypotension. However, the relative impact of these factors remains unclear.

To address this, we analysed 399 post-mortem brains from LBD cases comprising Parkinson’s disease, Parkinson’s disease dementia and dementia with Lewy bodies, and controls, integrating APOE genotype, clinical data and assessment of ischaemic pathology alongside large-scale digital pathology quantification. We established an image analysis pipeline using machine learning to enable automated, standardized measurement of α-synuclein, amyloid-β and phosphorylated tau burden across multiple brain regions.

Quantitative pathology was strongly correlated with semi-quantitative ratings and outperformed conventional staging in predicting dementia. Across multiple analytical approaches, APOE ɛ3 and ɛ4 carriers showed distinct dementia risk profiles. APOE ɛ3 carriers developed dementia at lower quantitative α-synuclein and amyloid-β thresholds than ɛ4 carriers, although overall dementia risk was dominated by ɛ4 genotype, consistent with ɛ4 both promoting greater pathology accumulation and modifying the threshold for dementia onset. Orthostatic hypotension and ischaemic pathology increased dementia risk only in ɛ3 carriers with low Lewy and Alzheimer’s proteinopathy burden, and male sex further modulated dementia risk for this subgroup. The Subtype and Stage Inference (SuStaIn) algorithm identified four trajectories of Lewy pathology progression. Two corresponded to recognized patterns: (i) brainstem first; and (ii) with early amygdala and concomitant brainstem involvement. Two further patterns showed early cortical involvement: (i) with early cingulate cortex involvement together with brainstem pathology; and (ii) starting in the neocortex before limbic and brainstem involvement. Co-pathology progression modelling identified subtypes with early predominance of amyloid-β, phosphorylated tau or α-synuclein and showed that Lewy subtypes follow two propagation trajectories in opposite directions.

Together, these findings demonstrate that integrating quantitative pathology with genotype and clinical data reveals distinct yet overlapping pathways to dementia in LBD, refining disease progression models and providing a basis for genotype- and pathology-informed patient stratification in therapeutic trials.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 7 matches between paragraphs and lines of code.

Zenodo 18315069

License: CC-BY-4.0

State: the link answers, verified on 4 November 2026

Evidence: files inventoried

Languages: R (4), Jupyter (2)

Size: 12 files, 6 scripts

Software Heritage: not checked

Found in: “Data availability”

Holds: 2 notebooks

Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation

Tools: tidyverse (4 files), ggplot2 (3 files), Matplotlib (2 files), NumPy (2 files), pandas (2 files), Plotly (2 files), scikit-learn (2 files), scikit-posthocs (2 files), SciPy (2 files), seaborn (2 files), statsmodels (2 files), car (1 file), cowplot (1 file), pROC (1 file), reshape2 (1 file), rstatix (1 file), statannotations (1 file)

Availability: 1 check, the latest on 4 November 2026: the link answers (HTTP 200)

4 November 2026: the link answers (HTTP 200)

6 files

QuantPath_Figure_1_and_2.R, R, 5,374 lines, 2 matches
QuantPath_Figure_3_and_Suppl_Fig_1.R, R, 1,022 lines
QuantPath_Figure_4_and_5.R, R, 728 lines, 2 matches
QuantPath_Table1_and_Suppl_Tables_3_4_5.R, R, 1,009 lines, 2 matches
SuStaIn_ASyn_ABeta_pTau_Subtype.ipynb, Jupyter, 2,051 lines
SuStaIn_ASyn_Subtype.ipynb, Jupyter, 1,935 lines, 1 match

At the source: zenodo.org/records/18315069

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
6 scripts, each with its path and the digest of its content;
7 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability

The raw quantitative pathological data, protocols, R and python code and key laboratory materials used and generated in this study are listed in a Key Resource Table (KRT) alongside their persistent identifiers at https://zenodo.org/records/18315069. Whole slide images of cases used in this study are available for viewing via UCL QSBB Digital Pathology Resource (RRID:SCR_025020). Anonymized clinical and pathology data are available upon reasonable request, if in compliance with regulatory ethical approval.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 4 November 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 19 authors, 6 keywords, 16 MeSH terms, 2 funders, 0 references, 1 RRID.

Cite

This paper

Nelvagal, H. R., Chiraki, N., Curless, T., Cullinane, P. W., Rockliffe, A., Pimparkar, S., Kawamura, H., Ollerenshaw, S., Elahi, I., Brandner, S., Wu, L., Real, R., Ryten, M., Hardy, J., De Pablo Fernandez, E., Warner, T. T., Morris, H. R., Lim, Y. M., & Jaunmuktane, Z. (2026). Quantitative pathology and APOE genotype reveal dementia risk and progression in Lewy body disease. Brain : a journal of neurology, 149(11), 3905-3922. https://doi.org/10.1093/brain/awag114

BibTeX

@article{nelvagal2026quantitative, author = {Nelvagal, Hemanth R and Chiraki, Nancy and Curless, Toby and Cullinane, Patrick W and Rockliffe, Alice and Pimparkar, Samarth and Kawamura, Haruka and Ollerenshaw, Sophie and Elahi, Isha and Brandner, Sebastian and Wu, Lesley and Real, Raquel and Ryten, Mina and Hardy, John and De Pablo Fernandez, Eduardo and Warner, Thomas T and Morris, Huw R and Lim, Yau Mun and Jaunmuktane, Zane}, title = {{Quantitative pathology and APOE genotype reveal dementia risk and progression in Lewy body disease}}, journal = {Brain : a journal of neurology}, year = {2026}, month = nov, volume = {149}, number = {11}, pages = {3905--3922}, publisher = {Oxford University Press}, issn = {0006-8950}, doi = {10.1093/brain/awag114}, url = {https://doi.org/10.1093/brain/awag114}, pmid = {41889331}, pmcid = {PMC13730591} }

RIS

TY - JOUR AU - Nelvagal, Hemanth R AU - Chiraki, Nancy AU - Curless, Toby AU - Cullinane, Patrick W AU - Rockliffe, Alice AU - Pimparkar, Samarth AU - Kawamura, Haruka AU - Ollerenshaw, Sophie AU - Elahi, Isha AU - Brandner, Sebastian AU - Wu, Lesley AU - Real, Raquel AU - Ryten, Mina AU - Hardy, John AU - De Pablo Fernandez, Eduardo AU - Warner, Thomas T AU - Morris, Huw R AU - Lim, Yau Mun AU - Jaunmuktane, Zane TI - Quantitative pathology and APOE genotype reveal dementia risk and progression in Lewy body disease T2 - Brain : a journal of neurology J2 - Brain PY - 2026 DA - 2026/11/01 VL - 149 IS - 11 SP - 3905 EP - 3922 SN - 0006-8950 PB - Oxford University Press DO - 10.1093/brain/awag114 UR - https://doi.org/10.1093/brain/awag114 LA - en ER -

CSL-JSON

{ "id": "10.1093/brain/awag114", "type": "article-journal", "title": "Quantitative pathology and APOE genotype reveal dementia risk and progression in Lewy body disease", "container-title": "Brain : a journal of neurology", "author": [ { "family": "Nelvagal", "given": "Hemanth R" }, { "family": "Chiraki", "given": "Nancy" }, { "family": "Curless", "given": "Toby" }, { "family": "Cullinane", "given": "Patrick W" }, { "family": "Rockliffe", "given": "Alice" }, { "family": "Pimparkar", "given": "Samarth" }, { "family": "Kawamura", "given": "Haruka" }, { "family": "Ollerenshaw", "given": "Sophie" }, { "family": "Elahi", "given": "Isha" }, { "family": "Brandner", "given": "Sebastian" }, { "family": "Wu", "given": "Lesley" }, { "family": "Real", "given": "Raquel" }, { "family": "Ryten", "given": "Mina" }, { "family": "Hardy", "given": "John" }, { "family": "De Pablo Fernandez", "given": "Eduardo" }, { "family": "Warner", "given": "Thomas T" }, { "family": "Morris", "given": "Huw R" }, { "family": "Lim", "given": "Yau Mun" }, { "family": "Jaunmuktane", "given": "Zane" } ], "container-title-short": "Brain", "volume": "149", "issue": "11", "page": "3905-3922", "DOI": "10.1093/brain/awag114", "PMID": "41889331", "PMCID": "PMC13730591", "ISSN": "0006-8950", "publisher": "Oxford University Press", "URL": "https://doi.org/10.1093/brain/awag114", "language": "en", "issued": { "date-parts": [ [ 2026, 11, 1 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1093/brain/awag131 [code]: Data-driven modelling of tau pathology reveals distinct progressive supranuclear palsy subtypes.
Journal: Brain : a journal of neurology
In common: Alzheimer's / dementia, 9 authors
[2] doi:10.1212/nxg.0000000000200092 [code]: Genome-wide Analysis of Motor Progression in Parkinson Disease.
Journal: Neurology. Genetics
In common: Plotly, reshape2, ggplot2, 4 other tools, Parkinson's, genetics / omics, 4 authors
[3] doi:10.1038/s41467-024-52025-x [code]: Genetic, transcriptomic, histological, and biochemical analysis of progressive supranuclear palsy implicates glial activation and novel risk genes.
Journal: Nature communications
In common: tidyverse, pandas, Matplotlib, 1 other tool, Alzheimer's / dementia, genetics / omics, 4 authors
[4] doi:10.1126/sciadv.adk1296 [code]: The annotation of <i>GBA1</i> has been concealed by its protein-coding pseudogene <i>GBAP1</i>.
Journal: Science advances
In common: rstatix, car, cowplot, 7 other tools, Parkinson's, genetics / omics, 2 authors
[5] doi:10.21203/rs.3.rs-8766017/v1 [code]: A molecular map of the living human brain from quantitative MRI
Journal: Research Square (preprint)
In common: reshape2, statsmodels, ggplot2, 7 other tools, genetics / omics, 2 authors
[6] doi:10.1126/sciadv.adp8504 [code]: Astrocytic RNA editing regulates the host immune response to alpha-synuclein.
Journal: Science advances
In common: cowplot, reshape2, ggplot2, 4 other tools, Parkinson's, 2 authors
[7] doi:10.1038/s41588-023-01300-6 [code]: Brain expression quantitative trait locus and network analyses reveal downstream effects and putative drivers for brain-related diseases.
Journal: Nature genetics
In common: pROC, statannotations, Plotly, 11 other tools, genetics / omics
[8] doi:10.1093/bioinformatics/btag592 [code]: Network-based stratification of allele-specific expression reveals patient subgroups in Huntington's disease.
Journal: Bioinformatics (Oxford, England)
In common: pROC, rstatix, car, 11 other tools, genetics / omics
[9] doi:10.1016/j.isci.2023.107522 [code]: Deep learning for risk-based stratification of cognitively impaired individuals.
Journal: iScience
In common: scikit-posthocs, rstatix, Plotly, 10 other tools, Alzheimer's / dementia
[10] doi:10.1038/s41467-025-66783-9 [code]: Evidence for divergent cortical organisation in Parkinson's disease and Lewy Body Dementia.
Journal: Nature communications
In common: scikit-posthocs, statsmodels, seaborn, 5 other tools, Parkinson's, Alzheimer's / dementia, author Mina Ryten

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Add the badge to the code's README

If you are an author or a maintainer of this paper's code, you can add this badge. It says what OSCR knows of the paper (its code, its validated tracing map) and links the code to this page. Paste one of these snippets under the README's title. Keep the two marks around it, so that the badge can be replaced in place and is never added twice. OSCR changes nothing in your repository.

Markdown

<!-- oscr-badge:start -->
[![OSCR](https://openscicode.org/badge/doi/10.1093/brain/awag114.svg)](https://openscicode.org/badge/doi/10.1093/brain/awag114)
<!-- oscr-badge:end -->

reStructuredText

.. oscr-badge:start

.. image:: https://openscicode.org/badge/doi/10.1093/brain/awag114.svg
   :target: https://openscicode.org/badge/doi/10.1093/brain/awag114
   :alt: OSCR

.. oscr-badge:end

HTML

<!-- oscr-badge:start -->
<a href="https://openscicode.org/badge/doi/10.1093/brain/awag114"><img src="https://openscicode.org/badge/doi/10.1093/brain/awag114.svg" alt="OSCR"></a>
<!-- oscr-badge:end -->

If you choose to, you can have the badge kept up to date for you, with oscr badge --write from the command line or with the GitHub Action in your repository's workflows. Every other form and option (a card, AsciiDoc, LaTeX, shields.io) is on the badge page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: code errors and mismatches between the code and the paper about this paper and its code, from readers and its authors, beside the paper and its tracing map.

Report a mismatch between the code and the paper · Report a code error (signed in)

Reproductions: reports from readers who ran the authors' code and did not get the paper's result, what they ran, with which environment, commit and data.

Report a reproduction failure (signed in)

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.