OSCR

Glial and Vascular Plasma Biomarkers Across the Alzheimer's Disease Continuum: An ADNI-Based Longitudinal Study.

Code ↔ Paper

17 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 17 matches
  1. [1] § Materials and Methods › Biomarker Harmonization and Transformation ↔ 13_revision_diagnostics.R, lines 112–158 · score 0.83 · regression slope, natural log, raw multiplex, sVCAM, sICAM, pg
  2. [2] § Materials and Methods › Study Design and Data Source ↔ 06_merge_selected_biomarkers.R, lines 146–185 · score 0.80 · intercellular adhesion molecule, vascular endothelial growth, vascular cell adhesion, Plasma, biomarker
  3. [3] § Materials and Methods › Sensitivity Analyses ↔ 15_sensitivity_analyses.R, lines 309–389 · score 0.77 · informative dropout, stable MCI, AD converters, baseline biomarker, split, predicted
  4. [4] § Materials and Methods › Study Design and Data Source ↔ 13_revision_diagnostics.R, lines 160–246 · score 0.74 · vascular endothelial growth, vascular cell adhesion, intercellular adhesion, protein, Plasma
  5. [5] § Results › Sensitivity Analyses ↔ 15_sensitivity_analyses.R, lines 309–389 · score 0.70 · conversion rates, MCI converted, stable MCI, AD converters, analytic, Sensitivity
  6. [6] § Materials and Methods › Biomarker Harmonization and Transformation ↔ 15_sensitivity_analyses.R, lines 91–130 · score 0.69 · Myriad RBM, Biomarkers Consortium, sICAM, vascular
  7. [7] § Materials and Methods › Biomarker Harmonization and Transformation ↔ 14_corrected_primary_analysis.R, lines 50–100 · score 0.67 · sVCAM, sTREM2, sICAM, pg, log10, coefficients
  8. [8] § Results › Baseline Glial‐Vascular Correlation Structure ↔ 07_baseline_and_longitudinal_analysis.R, lines 336–425 · score 0.66 · correlation heatmap, Baseline glial vascular, Spearman correlations, Tile, log, diagnostic
  9. [9] § Materials and Methods › Vascular Risk Factor Adjustment ↔ 18_risk_adjustment_repaired.R, lines 350–421 · score 0.63 · renal function, blood pressure, antidiabetic, antihypertensive, smoking, Risk
  10. [10] § Materials and Methods › Biomarker Harmonization and Transformation ↔ 05_biomarker_candidate_QA.R, lines 40–107 · score 0.63 · Biomarkers Consortium, Haass, Proteomics, WashU, sTREM2, Quanterix
  11. [11] § Results › Baseline Biomarker Distributions ↔ 14_corrected_primary_analysis.R, lines 50–100 · score 0.63 · plasma GFAP, sVCAM, sTREM2, sICAM, ng, pg
  12. [12] § Materials and Methods › Baseline Analysis ↔ 14_corrected_primary_analysis.R, lines 355–406 · score 0.59 · overlap coefficient, Cohen, unadjusted, detectable, discriminative, power
  13. [13] § Results › Sensitivity Analyses ↔ 17_fixes_and_risk_covariates (1).R, lines 514–593 · score 0.58 · Vascular risk, vascular models, unadjusted, antidiabetic, antihypertensive, smoking
  14. [14] § Materials and Methods › Data Extraction and Cohort Construction ↔ 15_sensitivity_analyses.R, lines 243–307 · score 0.58 · conversion status, varying diagnosis, sensitivity, date, RID, MCI
  15. [15] § Materials and Methods › Correlation Analysis ↔ 07_baseline_and_longitudinal_analysis.R, lines 336–425 · score 0.56 · Baseline glial vascular, Spearman correlations, log, biomarker
  16. [16] § Materials and Methods › Multiple‐Comparison Correction ↔ 13_revision_diagnostics.R, lines 501–555 · score 0.52 · reported baseline, definitions, families, coefficient, intercept, sensitivity
  17. [17] § Materials and Methods › Longitudinal Analysis ↔ 14_corrected_primary_analysis.R, lines 230–290 · score 0.50 · random slope, random intercept, Residual, visits, vascular, model

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

R · 389 lines · 17 KB · MIT · 4 matches

  1. # ==========================================================
  2. # 15_sensitivity_analyses.R
  3. #
  4. # Sensitivity analyses requested by Reviewer 2 (comments 2, 3),
  5. # plus a batch-effect investigation of the vascular longitudinal
  6. # finding that neither reviewer raised but that determines how
  7. # that finding should be framed.
  8. #
  9. # Sections
  10. # A. Vascular batch / assay-run investigation
  11. # B. Restriction to participants with 2+ measurements (R2-2)
  12. # C. Restriction to 2+ years of follow-up (R2-2)
  13. # D. MCI converters and time-varying diagnosis (R2-3)
  14. # E. Informative dropout probe
  15. #
  16. # Run 14_corrected_primary_analysis.R first; this script repeats
  17. # the corrected outcome construction so it can stand alone.
  18. # ==========================================================
  20. needed <- c("tidyverse", "janitor", "openxlsx", "lme4", "lmerTest",
  21. "broom", "broom.mixed")
  22. missing <- needed[!needed %in% rownames(installed.packages())]
  23. if (length(missing) > 0) install.packages(missing)
  25. library(tidyverse); library(janitor); library(openxlsx)
  26. library(lme4); library(lmerTest); library(broom); library(broom.mixed)
  28. project_dir <- "R:/ADNI_Project"
  29. bio_dir <- file.path(project_dir, "00_raw_data", "biomarkers_excel")
  30. clean_dir <- file.path(project_dir, "02_clean_data")
  31. results_dir <- file.path(project_dir, "04_results")
  33. out <- list()
  34. parse_num <- function(x) readr::parse_number(as.character(x))
  35. first_nonmissing <- function(x) { x <- x[!is.na(x) & x != ""]; if (!length(x)) NA else x[1] }
  37. marker_info <- tribble(
  38. ~marker, ~label, ~stored_scale,
  39. "gfap_quanterix", "Plasma GFAP", "raw",
  40. "strem2_msd_corrected", "Plasma sTREM2", "raw",
  41. "vegf_plasma_qc", "Plasma VEGF", "log10",
  42. "sicam1_plasma_qc", "Plasma sICAM-1", "log10",
  43. "svcam1_plasma_qc", "Plasma sVCAM-1", "log10"
  44. )
  45. primary_markers <- marker_info$marker
  46. vascular_markers <- c("vegf_plasma_qc", "sicam1_plasma_qc", "svcam1_plasma_qc")
  48. dat <- read_csv(file.path(clean_dir, "analysis_master_model_ready.csv"),
  49. show_col_types = FALSE, guess_max = 100000) %>%
  50. clean_names() %>%
  51. mutate(
  52. rid = as.character(rid), visit_key = as.character(visit_key),
  53. dx_label = factor(dx_label, levels = c("CN", "MCI", "AD")),
  54. baseline_dx = factor(baseline_dx, levels = c("CN", "MCI", "AD")),
  55. ptgender = factor(ptgender), age = parse_num(age),
  56. pteducat = parse_num(pteducat), apoe4 = parse_num(apoe4),
  57. years_from_baseline = parse_num(years_from_baseline)
  58. ) %>%
  59. mutate(baseline_age = age - years_from_baseline)
  61. for (i in seq_len(nrow(marker_info))) {
  62. m <- marker_info$marker[i]; v <- parse_num(dat[[m]])
  63. dat[[paste0("ln_", m)]] <- if (marker_info$stored_scale[i] == "log10")
  64. log(10) * v else ifelse(!is.na(v) & v > 0, log(v), NA_real_)
  65. }
  67. long_cc <- function(m, d = dat) {
  68. d %>% filter(!is.na(.data[[paste0("ln_", m)]]), !is.na(years_from_baseline),
  69. !is.na(baseline_dx), !is.na(baseline_age), !is.na(ptgender),
  70. !is.na(pteducat), !is.na(apoe4))
  71. }
  73. reported_long <- c("years_from_baseline", "baseline_dxMCI", "baseline_dxAD",
  74. "years_from_baseline:baseline_dxMCI", "years_from_baseline:baseline_dxAD")
  76. fit_long <- function(m, d) {
  77. lmer(as.formula(paste0("ln_", m,
  78. " ~ years_from_baseline * baseline_dx + baseline_age + ptgender + pteducat + apoe4 + (1 | rid)")),
  79. data = d, REML = FALSE)
  80. }
  82. extract_long <- function(fit, m, tag) {
  83. tidy(fit, effects = "fixed") %>%
  84. filter(term %in% reported_long) %>%
  85. mutate(marker = m, analysis = tag,
  86. percent_change = (exp(estimate) - 1) * 100,
  87. n_obs = nobs(fit), n_subj = ngrps(fit)[["rid"]])
  88. }
  91. # ==========================================================
  92. # A. VASCULAR BATCH / ASSAY-RUN INVESTIGATION
  93. #
  94. # The vascular panel has exactly two visits, bl and m12, assayed
  95. # by Myriad RBM as part of the Biomarkers Consortium project.
  96. # The published sICAM-1 CN slope corresponds to roughly a 24%
  97. # fall in concentration over 12 months, which is not a plausible
  98. # biological rate. Three checks distinguish biology from a
  99. # run-to-run shift.
  100. # ==========================================================
  102. cat("\n=== A. Vascular batch investigation ===\n")
  104. # A1. Paired within-person change from bl to m12, by diagnosis.
  105. # A uniform shift across all three groups points to a batch
  106. # effect; a shift confined to one group points to biology.
  108. paired_change <- map_dfr(vascular_markers, function(m) {
  109. ln <- paste0("ln_", m)
  110. dat %>%
  111. filter(visit_key %in% c("bl", "m12"), !is.na(.data[[ln]]), !is.na(baseline_dx)) %>%
  112. select(rid, baseline_dx, visit_key, value = all_of(ln)) %>%
  113. pivot_wider(names_from = visit_key, values_from = value) %>%
  114. filter(!is.na(bl), !is.na(m12)) %>%
  115. mutate(delta = m12 - bl) %>%
  116. group_by(baseline_dx) %>%
  117. summarise(
  118. n = n(),
  119. mean_delta_log = mean(delta),
  120. percent_change_12m = (exp(mean(delta)) - 1) * 100,
  121. sd_delta = sd(delta),
  122. t_p = t.test(delta)$p.value,
  123. pct_declining = round(100 * mean(delta < 0), 1),
  124. .groups = "drop"
  125. ) %>%
  126. mutate(marker = m)
  127. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  129. print(as.data.frame(paired_change))
  130. out$A1_paired_change <- paired_change
  132. # A2. Is the bl-to-m12 shift the same in every group?
  133. # If the group differences vanish, the shift is common to all
  134. # samples and behaves like a run effect.
  135. out$A2_shift_homogeneity <- map_dfr(vascular_markers, function(m) {
  136. ln <- paste0("ln_", m)
  137. d <- dat %>%
  138. filter(visit_key %in% c("bl", "m12"), !is.na(.data[[ln]]), !is.na(baseline_dx)) %>%
  139. select(rid, baseline_dx, visit_key, value = all_of(ln)) %>%
  140. pivot_wider(names_from = visit_key, values_from = value) %>%
  141. filter(!is.na(bl), !is.na(m12)) %>%
  142. mutate(delta = m12 - bl)
  143. a <- aov(delta ~ baseline_dx, data = d)
  144. tibble(marker = m,
  145. overall_mean_shift_pct = (exp(mean(d$delta)) - 1) * 100,
  146. anova_p_group_difference = summary(a)[[1]][["Pr(>F)"]][1],
  147. interpretation = if (summary(a)[[1]][["Pr(>F)"]][1] > 0.05)
  148. "shift is uniform across groups -> consistent with a run effect"
  149. else "shift differs by group -> not explained by a uniform run effect")
  150. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  152. # A3. Independent source check. ADMC_CLINICALVARIABLES is a separate
  153. # processing of the same three analytes. If the 12-month drop
  154. # appears there too it is in the samples, not the QC pipeline.
  156. admc_path <- file.path(bio_dir, "ADMC_CLINICALVARIABLES_16May2016.csv")
  157. if (file.exists(admc_path)) {
  158. admc <- read_csv(admc_path, show_col_types = FALSE, guess_max = 100000) %>% clean_names()
  159. key_visit <- intersect(c("viscode2", "viscode"), names(admc))[1]
  161. admc_long <- admc %>%
  162. transmute(rid = as.character(rid), visit_key = as.character(.data[[key_visit]]),
  163. vegf_plasma_qc = parse_num(vegf),
  164. sicam1_plasma_qc = parse_num(icam),
  165. svcam1_plasma_qc = parse_num(vcam)) %>%
  166. group_by(rid, visit_key) %>%
  167. summarise(across(everything(), first_nonmissing), .groups = "drop") %>%
  168. mutate(across(all_of(vascular_markers), as.numeric)) %>%
  169. pivot_longer(all_of(vascular_markers), names_to = "marker", values_to = "admc_value")
  171. out$A3_admc_check <- admc_long %>%
  172. filter(visit_key %in% c("bl", "m12"), !is.na(admc_value), admc_value > 0) %>%
  173. left_join(dat %>% distinct(rid, baseline_dx), by = "rid") %>%
  174. mutate(ln_admc = log(admc_value)) %>%
  175. select(rid, marker, visit_key, ln_admc, baseline_dx) %>%
  176. pivot_wider(names_from = visit_key, values_from = ln_admc) %>%
  177. filter(!is.na(bl), !is.na(m12)) %>%
  178. group_by(marker) %>%
  179. summarise(n = n(), mean_delta_log = mean(m12 - bl),
  180. percent_change_12m = (exp(mean(m12 - bl)) - 1) * 100,
  181. t_p = t.test(m12 - bl)$p.value, .groups = "drop") %>%
  182. mutate(note = "compare percent_change_12m with sheet A1; agreement means the shift is in the samples")
  183. print(as.data.frame(out$A3_admc_check))
  184. } else {
  185. cat("ADMC file not found, skipping the independent source check.\n")
  186. }
  188. # A4. Does the raw multiplex release carry plate / run / batch fields?
  189. raw_path <- file.path(bio_dir, "adni_plasma_raw_multiplex_11Nov2010.csv")
  190. if (file.exists(raw_path)) {
  191. raw_mx <- read_csv(raw_path, show_col_types = FALSE, guess_max = 100000) %>% clean_names()
  192. batch_cols <- grep("plate|batch|run|assay|date|lot|kit", names(raw_mx),
  193. value = TRUE, ignore.case = TRUE)
  194. out$A4_batch_fields <- tibble(
  195. candidate_batch_columns = if (length(batch_cols)) paste(batch_cols, collapse = ", ")
  196. else "none found",
  197. all_columns = paste(names(raw_mx), collapse = " | ")
  198. )
  199. cat("Candidate batch fields in raw multiplex file:",
  200. if (length(batch_cols)) paste(batch_cols, collapse = ", ") else "none", "\n")
  201. }
  204. # ==========================================================
  205. # B. RESTRICTION TO PARTICIPANTS WITH 2+ MEASUREMENTS
  206. #
  207. # Participants with a single observation contribute nothing to
  208. # within-person slope estimation but do pull the pooled time
  209. # coefficient toward a between-person contrast.
  210. # ==========================================================
  212. cat("\n=== B. Two-or-more measurements ===\n")
  214. sens_2plus <- map_dfr(primary_markers, function(m) {
  215. d <- long_cc(m) %>% group_by(rid) %>% filter(n() >= 2) %>% ungroup()
  216. if (n_distinct(d$rid) < 30 || n_distinct(d$baseline_dx) < 2) return(NULL)
  217. extract_long(fit_long(m, d), m, "restricted to 2+ measurements")
  218. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  219. out$B_two_or_more_visits <- sens_2plus
  221. # Within-between (Mundlak) decomposition, which uses the whole sample
  222. # but separates within-person change from between-person differences.
  223. mundlak <- map_dfr(primary_markers, function(m) {
  224. d <- long_cc(m) %>%
  225. group_by(rid) %>%
  226. mutate(time_mean = mean(years_from_baseline),
  227. time_within = years_from_baseline - time_mean) %>%
  228. ungroup()
  229. if (sd(d$time_within, na.rm = TRUE) == 0) return(NULL)
  230. fit <- lmer(as.formula(paste0("ln_", m,
  231. " ~ time_within * baseline_dx + time_mean + baseline_age + ptgender +",
  232. " pteducat + apoe4 + (1 | rid)")), data = d, REML = FALSE)
  233. tidy(fit, effects = "fixed") %>%
  234. filter(grepl("time_within|time_mean", term)) %>%
  235. mutate(marker = m, percent_change = (exp(estimate) - 1) * 100,
  236. reading = if_else(grepl("time_mean", term),
  237. "between-person component",
  238. "within-person component"))
  239. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  240. out$B_within_between <- mundlak
  243. # ==========================================================
  244. # C. RESTRICTION TO 2+ YEARS OF FOLLOW-UP
  245. # Not possible for the vascular panel, whose maximum follow-up
  246. # is about 1.6 years. That fact is itself part of the answer.
  247. # ==========================================================
  249. cat("\n=== C. Two-or-more years of follow-up ===\n")
  251. followup_capacity <- map_dfr(primary_markers, function(m) {
  252. d <- long_cc(m) %>% group_by(rid) %>%
  253. summarise(span = max(years_from_baseline) - min(years_from_baseline), .groups = "drop")
  254. tibble(marker = m, n_participants = nrow(d),
  255. n_with_2y_span = sum(d$span >= 2),
  256. pct_with_2y_span = round(100 * mean(d$span >= 2), 1),
  257. max_span = round(max(d$span), 2))
  258. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  259. out$C_followup_capacity <- followup_capacity
  261. sens_2y <- map_dfr(primary_markers, function(m) {
  262. ids <- long_cc(m) %>% group_by(rid) %>%
  263. summarise(span = max(years_from_baseline) - min(years_from_baseline), .groups = "drop") %>%
  264. filter(span >= 2) %>% pull(rid)
  265. d <- long_cc(m) %>% filter(rid %in% ids)
  266. if (n_distinct(d$rid) < 30 || n_distinct(d$baseline_dx) < 2) return(NULL)
  267. extract_long(fit_long(m, d), m, "restricted to 2+ years of follow-up")
  268. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  269. out$C_two_year_followup <- sens_2y
  272. # ==========================================================
  273. # D. MCI CONVERTERS AND TIME-VARYING DIAGNOSIS (Reviewer 2, comment 3)
  274. #
  275. # dx_label is already visit-level, so conversion status and a
  276. # time-varying diagnosis model are both available without new data.
  277. # ==========================================================
  279. cat("\n=== D. Converters ===\n")
  281. conversion <- dat %>%
  282. filter(!is.na(dx_label), !is.na(baseline_dx), !is.na(years_from_baseline)) %>%
  283. arrange(rid, years_from_baseline) %>%
  284. group_by(rid, baseline_dx) %>%
  285. summarise(
  286. n_visits_with_dx = n(),
  287. ever_ad = any(dx_label == "AD"),
  288. ever_mci = any(dx_label == "MCI"),
  289. last_dx = last(dx_label),
  290. time_to_first_ad = { i <- which(dx_label == "AD")
  291. if (length(i)) years_from_baseline[i[1]] else NA_real_ },
  292. followup_span = max(years_from_baseline) - min(years_from_baseline),
  293. .groups = "drop"
  294. ) %>%
  295. mutate(converter = case_when(
  296. baseline_dx == "MCI" & ever_ad ~ "MCI to AD converter",
  297. baseline_dx == "MCI" ~ "stable MCI",
  298. baseline_dx == "CN" & (ever_mci | ever_ad) ~ "CN progressor",
  299. baseline_dx == "CN" ~ "stable CN",
  300. TRUE ~ "AD at baseline"
  301. ))
  303. conversion_summary <- conversion %>%
  304. count(baseline_dx, converter) %>%
  305. group_by(baseline_dx) %>% mutate(percent = round(100 * n / sum(n), 1)) %>% ungroup()
  306. print(as.data.frame(conversion_summary))
  307. out$D1_conversion_summary <- conversion_summary
  309. # Conversion rates within each biomarker analytic sample
  310. out$D2_conversion_by_panel <- map_dfr(primary_markers, function(m) {
  311. ids <- unique(long_cc(m)$rid)
  312. conversion %>% filter(rid %in% ids) %>% count(marker = m, baseline_dx, converter) %>%
  313. group_by(marker, baseline_dx) %>% mutate(percent = round(100 * n / sum(n), 1)) %>% ungroup()
  314. })
  316. # D3. Time-varying diagnosis model
  317. out$D3_time_varying_dx <- map_dfr(primary_markers, function(m) {
  318. d <- long_cc(m) %>% filter(!is.na(dx_label))
  319. if (n_distinct(d$rid) < 30) return(NULL)
  320. fit <- lmer(as.formula(paste0("ln_", m,
  321. " ~ years_from_baseline * dx_label + baseline_age + ptgender + pteducat + apoe4 + (1 | rid)")),
  322. data = d, REML = FALSE)
  323. tidy(fit, effects = "fixed") %>%
  324. filter(grepl("years_from_baseline|dx_label", term)) %>%
  325. mutate(marker = m, analysis = "time-varying diagnosis",
  326. percent_change = (exp(estimate) - 1) * 100,
  327. n_obs = nobs(fit), n_subj = ngrps(fit)[["rid"]])
  328. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  330. # D4. Split the MCI arm into stable and converter
  331. out$D4_mci_split <- map_dfr(primary_markers, function(m) {
  332. d <- long_cc(m) %>%
  333. left_join(conversion %>% select(rid, converter), by = "rid") %>%
  334. mutate(dx_group = factor(case_when(
  335. baseline_dx == "CN" ~ "CN",
  336. converter == "stable MCI" ~ "stable MCI",
  337. converter == "MCI to AD converter" ~ "MCI converter",
  338. baseline_dx == "AD" ~ "AD",
  339. TRUE ~ NA_character_),
  340. levels = c("CN", "stable MCI", "MCI converter", "AD"))) %>%
  341. filter(!is.na(dx_group))
  342. if (n_distinct(d$dx_group) < 3 || n_distinct(d$rid) < 40) return(NULL)
  343. fit <- lmer(as.formula(paste0("ln_", m,
  344. " ~ years_from_baseline * dx_group + baseline_age + ptgender + pteducat + apoe4 + (1 | rid)")),
  345. data = d, REML = FALSE)
  346. tidy(fit, effects = "fixed") %>%
  347. filter(grepl("years_from_baseline|dx_group", term)) %>%
  348. mutate(marker = m, analysis = "MCI split by conversion",
  349. percent_change = (exp(estimate) - 1) * 100)
  350. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  353. # ==========================================================
  354. # E. INFORMATIVE DROPOUT PROBE
  355. #
  356. # Does the baseline biomarker value predict whether a participant
  357. # returns for a follow-up measurement? If it does, dropout is
  358. # informative and the slopes need that caveat.
  359. # ==========================================================
  361. cat("\n=== E. Informative dropout ===\n")
  363. dropout <- map_dfr(primary_markers, function(m) {
  364. ln <- paste0("ln_", m)
  365. d <- long_cc(m) %>% group_by(rid) %>%
  366. mutate(returned = as.integer(n() >= 2)) %>%
  367. filter(years_from_baseline == min(years_from_baseline)) %>%
  368. slice(1) %>% ungroup()
  369. if (n_distinct(d$returned) < 2) return(NULL)
  370. fit <- glm(returned ~ get(ln) + baseline_dx + baseline_age + ptgender + pteducat + apoe4,
  371. data = d, family = binomial)
  372. tidy(fit) %>% filter(grepl("get\\(ln\\)|baseline_dx", term)) %>%
  373. mutate(marker = m, odds_ratio = exp(estimate),
  374. term = dplyr::recode(term, `get(ln)` = "baseline biomarker level (log)"))
  375. }) %>% left_join(marker_info, by = "marker") %>% relocate(label)
  376. out$E_informative_dropout <- dropout
  379. out <- imap(out, function(x, nm) {
  380. if (is.null(x) || (is.data.frame(x) && (nrow(x) == 0 || ncol(x) == 0)))
  381. tibble(note = paste("No rows produced for", nm)) else x
  382. })
  384. # Excel caps sheet names at 31 characters.
  385. names(out) <- substr(make.unique(names(out)), 1, 31)
  387. openxlsx::write.xlsx(out, file.path(results_dir, "sensitivity_analyses.xlsx"),
  388. overwrite = TRUE)
  389. cat("\nDone. Results:", file.path(results_dir, "sensitivity_analyses.xlsx"), "\n")

15_sensitivity_analyses.R at commit 27e05ee, under MIT · at the source

Overview

Authors: Khodakaram Jahanbin1, Alzheimer's Disease Neuroimaging Initiative
ORCID iDs: Khodakaram Jahanbin
  1. Department of Immunology, School of Medicine Ahvaz Jundishapur University of Medical Sciences Ahvaz Iran
Institutions: Ahvaz Jundishapur University of Medical Sciences (Iran)
Journal: FASEB bioAdvances, volume 8, issue 9, article e70155
Dates: received 14 June 2026; accepted 14 September 2026; published online 19 September 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.1096/fba.2026-00194 · PMID 42763575 · PMCID PMC13589249 · OpenAlex W7213637425
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: other (modality), human (organism), Alzheimer's / dementia (population), clinical / translational (subfield)
Methods: Connectivity, Statistics
Keywords: Alzheimer's disease, biomarkers, blood–brain barrier, glial fibrillary acidic protein, vascular cell adhesion molecule‐1
Topic: Dementia and Cognitive Impairment Research (Psychiatry and Mental health, Medicine), according to OpenAlex
Funding: NIH (U01 AG024904); Department of Defense (W81XWH‐12‐2‐0012)
Citations: not cited yet (Europe PMC); 34 references in the paper

Abstract

Alzheimer's disease (AD) is increasingly recognized as a multicellular disorder involving neurovascular unit dysfunction. Investigating glial and vascular biomarkers together may provide a more integrated view of AD biology. This study characterized baseline distributions, interrelationships, and longitudinal trajectories of plasma glial (GFAP, sTREM2) and vascular/endothelial (VEGF, sICAM‐1, sVCAM‐1) biomarkers across the AD continuum. Using Alzheimer's Disease Neuroimaging Initiative (ADNI) data, this retrospective longitudinal cohort study included a covariate‐complete clinical cohort of 2650 participants classified as cognitively unimpaired (CU), mild cognitive impairment (MCI), or AD dementia. Biomarker‐specific analytic samples were determined by assay availability and complete‐case requirements. Associations were evaluated using covariate‐adjusted linear regression and linear mixed‐effects models with participant random intercepts, adjusted for baseline age, sex, education, and APOE ε4 status. Vascular models were additionally refitted with body mass index, systolic blood pressure, antihypertensive and antidiabetic medication use, smoking history, and estimated glomerular filtration rate. Baseline GFAP showed a robust stepwise elevation from CU to MCI to AD (56.3% higher in AD than CU, q = 1.7 × 10−14; area under the curve 0.82), whereas sTREM2 distributions overlapped across groups. VEGF and sVCAM‐1 were higher in AD than CU (14.1%, q = 0.014; 14.7%, q = 0.002), with areas under the curve of 0.61 and 0.63 and more than 80% distribution overlap. GFAP increased by 4.50% per year in CU participants and sTREM2 by 2.99% per year, with no significant diagnosis‐by‐time interactions for either. Over a 12‐month interval, sICAM‐1 declined in CU participants, and this decline was attenuated in MCI and AD, while sVCAM‐1 increased in CU participants and showed negative diagnosis‐by‐time interactions. The vascular longitudinal findings were unchanged by adjustment for cardiometabolic and renal covariates, by time‐varying diagnosis, and by separating stable MCI from MCI‐to‐AD converters. Correlations among vascular markers were consistent and well estimated, whereas glial‐vascular correlations were based on 64 to 94 overlapping participants and were not significant. Glial and vascular plasma biomarkers did not progress synchronously across the AD continuum. GFAP showed the strongest and most discriminating diagnosis‐associated elevation. Vascular markers showed statistically robust but small group‐level differences with no individual‐level discriminative utility, and their divergent 12‐month trajectories require replication over longer follow‐up before they can be interpreted as stage‐dependent regulation.

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 17 matches between paragraphs and lines of code.

Khodakaram/adni-glial-vascular-axis

License: MIT
State: the link answers, verified on 26 September 2026
Evidence: files inventoried
Commit: 27e05ee667a76f7ff19358c4fbb1bca2702607a3, 30 August 2026
Languages: R (15)
Size: 18 files, 15 scripts
Software Heritage: not archived
Found in: the text, “Study Design and Data Source”
Holds: README, license file
Not found: CITATION.cff, environment file, tests, continuous integration, documentation
Tools: tidyverse (15 files), lme4 (6 files), lmerTest (6 files), broom (5 files), car (1 file), pROC (1 file)
Availability: 1 check, the latest on 26 September 2026: the link answers
  • 26 September 2026: the link answers
17 files
At the source: github.com/Khodakaram/adni-glial-vascular-axis

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;
  • 15 scripts, each with its path and the digest of its content;
  • 17 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 Statement

The data used in this research were obtained from Alzheimer's Disease Neuroimaging Initiative (ADNI) and are available with permission to all researchers.

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, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 2 authors, 5 keywords, 2 funders, 32 references.

Cite

This paper

Jahanbin, K., & Alzheimer's Disease Neuroimaging Initiative. (2026). Glial and Vascular Plasma Biomarkers Across the Alzheimer's Disease Continuum: An ADNI-Based Longitudinal Study. FASEB bioAdvances, 8(9), e70155. https://doi.org/10.1096/fba.2026-00194

BibTeX

@article{jahanbin2026glial,
author = {Jahanbin, Khodakaram and {Alzheimer's Disease Neuroimaging Initiative}},
title = {{Glial and Vascular Plasma Biomarkers Across the Alzheimer's Disease Continuum: An ADNI-Based Longitudinal Study}},
journal = {FASEB bioAdvances},
year = {2026},
month = sep,
volume = {8},
number = {9},
pages = {e70155},
publisher = {Wiley},
issn = {2573-9832},
doi = {10.1096/fba.2026-00194},
url = {https://doi.org/10.1096/fba.2026-00194},
pmid = {42763575},
pmcid = {PMC13589249}
}

RIS

TY - JOUR
AU - Jahanbin, Khodakaram
AU - Alzheimer's Disease Neuroimaging Initiative
TI - Glial and Vascular Plasma Biomarkers Across the Alzheimer's Disease Continuum: An ADNI-Based Longitudinal Study
T2 - FASEB bioAdvances
J2 - FASEB Bioadv
PY - 2026
DA - 2026/09/19
VL - 8
IS - 9
SP - e70155
SN - 2573-9832
PB - Wiley
DO - 10.1096/fba.2026-00194
UR - https://doi.org/10.1096/fba.2026-00194
LA - en
ER -

CSL-JSON

{
"id": "10.1096/fba.2026-00194",
"type": "article-journal",
"title": "Glial and Vascular Plasma Biomarkers Across the Alzheimer's Disease Continuum: An ADNI-Based Longitudinal Study",
"container-title": "FASEB bioAdvances",
"author": [
{
"family": "Jahanbin",
"given": "Khodakaram"
},
{
"literal": "Alzheimer's Disease Neuroimaging Initiative"
}
],
"container-title-short": "FASEB Bioadv",
"volume": "8",
"issue": "9",
"page": "e70155",
"DOI": "10.1096/fba.2026-00194",
"PMID": "42763575",
"PMCID": "PMC13589249",
"ISSN": "2573-9832",
"publisher": "Wiley",
"URL": "https://doi.org/10.1096/fba.2026-00194",
"language": "en",
"issued": {
"date-parts": [
[
2026,
9,
19
]
]
}
}

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:10.1016/j.celrep.2026.117505 [code]
Impaired spatial coding and neuronal hyperactivity in the medial entorhinal cortex of aged APP knock-in mice.
Journal: Cell reports
In common: car, broom, lmerTest, 2 other tools, Alzheimer's / dementia, 1 reference
[2] doi:10.1038/s42003-026-10282-0 [code]
Genetic risk of Alzheimer's disease is associated with loss of brain network segregation in midlife.
Journal: Communications biology
In common: car, broom, lmerTest, 2 other tools, Alzheimer's / dementia, clinical / translational
[3] doi:10.1038/s41588-026-02722-8 [code]
A multiancestry polygenic risk score for Alzheimer's disease is associated with cognitive decline and neuropathological hallmarks in diverse populations.
Journal: Nature genetics
In common: pROC, lmerTest, lme4, 1 other tool, Alzheimer's / dementia, 1 reference
[4] doi:10.1038/s41467-026-72934-3 [code]
Multi-focal ultrasound neuromodulation to the dorsal anterior cingulate cortex disrupts behavioural and neural pain processing.
Journal: Nature communications
In common: car, broom, lmerTest, 2 other tools, other
[5] doi:10.1126/sciadv.aec9291 [code]
Computational mechanisms of perception in autism revealed using games inspired by rodent operant tasks.
Journal: Science advances
In common: pROC, car, broom, 2 other tools
[6] 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, car, broom, 2 other tools
[7] doi:10.1093/ageing/afag263 [code]
Cardiometabolic medication exposures and cognitive outcomes in Alzheimer's disease.
Journal: Age and ageing
In common: car, broom, lmerTest, 2 other tools, Alzheimer's / dementia
[8] doi:10.1016/j.xcrm.2026.102766 [code]
A longitudinal single-cell and spatial multiomic atlas of pediatric high-grade glioma.
Journal: Cell reports. Medicine
In common: pROC, car, lmerTest, 2 other tools
[9] doi:10.1038/s41467-026-77170-3 [code]
DNA methylation profiling identifies long-range epigenetic silencing of clustered protocadherins as a key determinant of meningioma progression.
Journal: Nature communications
In common: pROC, broom, lmerTest, 2 other tools
[10] doi:10.1016/j.celrep.2026.117646 [code]
Medial entorhinal-hippocampal desynchronization parallels the emergence of memory impairment in a mouse model of Alzheimer's disease pathology.
Journal: Cell reports
In common: car, broom, lmerTest, 2 other tools, Alzheimer's / dementia

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.

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: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

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