options(warn = -1)
+# Load user-defined functions
+source("../R/dependencies.R")
+source("../R/jointRPCA.R")
+source("../R/jointRPCAuniversal.R")
+source("../R/jointOptspaceHelper.R")
+source("../R/jointOptspaceSolve.R")
+source("../R/optspaceHelper.R")
+source("../R/transformHelper.R")
+source("../R/transform.R")
+source("../R/maskValueOnly.R")
+source("../R/rpcaTableProcessing.R")
+source("../R/jointRPCAutils.R")
+library(dplyr, warn.conflicts = FALSE)
+
+#paths
+f_mgx <- "data_ibdmdb_raw/taxonomic_profiles_mgx.tsv" #metagenomics
+f_mtx <- "data_ibdmdb_raw/ecs_relab.tsv" #metatranscriptomics
+f_meta <- "data_ibdmdb_raw/hmp2_metadata_2018-08-20.csv" #HMP2/IBDMDB master metadata
+
+# Helper functions
+
+#read an IBDMDB-style TSV with commented header line(s) beginning with "#"
+read_ibdmdb_tsv <- function(path) {
+ stopifnot(file.exists(path))
+ first <- readLines(path, n = 200L, warn = FALSE)
+ comment_idx <- which(grepl("^#", first))
+ if (length(comment_idx) == 0L) stop("No commented header line found in: ", path)
+
+ header_line <- first[max(comment_idx)]
+ header_line <- sub("^#\\s*", "", header_line)
+ header_line <- sub("^\ufeff", "", header_line)
+ header_vec <- strsplit(header_line, "\t", fixed = TRUE)[[1]]
+ header_vec <- gsub('^"|"$', "", header_vec)
+
+ dt <- fread(path, skip = length(comment_idx), header = FALSE, sep = "\t", quote = "")
+ if (ncol(dt) != length(header_vec)) {
+ stop(sprintf("Header columns (%d) != data columns (%d) in %s",
+ length(header_vec), ncol(dt), path))
+ }
+ setnames(dt, header_vec)
+ dt
+}
+
+#convert data.table (first column = feature IDs) => numeric matrix (features x samples)
+to_matrix <- function(dt) {
+ rn <- dt[[1]]
+ mat <- as.matrix(dt[, -1, with = FALSE])
+ rownames(mat) <- rn
+ storage.mode(mat) <- "numeric"
+ mat[is.na(mat)] <- 0
+ mat
+}
+
+#make rownames unique and ensure numeric / finite entries
+dedup_rownames <- function(mat) {
+ stopifnot(!is.null(rownames(mat)))
+ rn <- rownames(mat)
+ rn[rn == "" | is.na(rn)] <- paste0("feat_", seq_len(sum(rn == "" | is.na(rn))))
+ rownames(mat) <- make.unique(as.character(rn), sep = "_")
+ mat
+}
+
+sanitize_matrix <- function(mat) {
+ mat <- as.matrix(mat)
+ storage.mode(mat) <- "numeric"
+ mat[!is.finite(mat)] <- 0
+ mat
+}
+
+
+#try to get an IBD vs non-IBD grouping from metadata
+make_group_from_meta <- function(meta_df, sample_ids) {
+ #always return a frame with sample_id and Group
+ out <- data.frame(sample_id = sample_ids, Group = NA_character_,
+ stringsAsFactors = FALSE)
+
+ if (is.null(meta_df) || !nrow(meta_df)) return(out)
+
+ #make sure we have a plain data.frame with original names preserved
+ meta_df <- as.data.frame(meta_df, stringsAsFactors = FALSE)
+ #standardize names (lowercase + trim) for matching
+ names(meta_df) <- tolower(trimws(names(meta_df)))
+
+ #candidate ID columns (lowercase)
+ id_candidates <- tolower(trimws(c(
+ "external id","external.id","external_id",
+ "sample id","sample.id","sample_id",
+ "mgx.sampleid","mtx.sampleid",
+ "stool sample id: tube a (etoh)","sample id: tube b (no preservative)",
+ "stool_id","wr id","wr_id","wrid"
+ )))
+ id_col <- intersect(id_candidates, names(meta_df))
+ if (!length(id_col)) return(out)
+
+ #pick the first ID column that overlaps enough with our sample IDs
+ chosen_id <- NULL
+ for (cand in id_col) {
+ x <- as.character(meta_df[[cand]])
+ hits <- sum(!is.na(x) & x %in% sample_ids)
+ if (hits >= max(10L, floor(length(sample_ids) * 0.05))) { chosen_id <- cand; break }
+ }
+ if (is.null(chosen_id)) return(out)
+
+ md <- meta_df[, c(chosen_id, setdiff(names(meta_df), chosen_id)), drop = FALSE]
+ names(md)[1] <- "sample_id"
+
+ #candidate diagnosis columns
+ diag_candidates <- tolower(trimws(c(
+ "diagnosis", "dx", "disease_status", "disease.status",
+ "ibd_status", "ibd.status", "ibd_subtype", "diagnosis_subtype"
+ )))
+ dcols <- intersect(diag_candidates, names(md))
+ if (!length(dcols)) return(out)
+
+ #try columns until one produces enough labels
+ for (col in dcols) {
+ x <- as.character(md[[col]])
+ if (!length(x) || length(x) != nrow(md)) next
+
+ grp <- rep(NA_character_, nrow(md))
+ is_ibd <- grepl("\\b(uc|cd|ibd)\\b", x, ignore.case = TRUE)
+ is_non <- grepl("^\\s*non", x, ignore.case = TRUE)
+ grp[is_ibd] <- "IBD"
+ grp[is_non & !is_ibd] <- "non-IBD"
+
+ n_lab <- sum(!is.na(grp))
+ n_cls <- length(unique(na.omit(grp)))
+ if (n_lab >= max(10L, floor(length(sample_ids) * 0.05)) && n_cls >= 2) {
+ md$Group <- grp
+ joined <- dplyr::left_join(
+ out,
+ md[, c("sample_id", "Group")],
+ by = "sample_id"
+ )
+ joined$Group <- factor(joined$Group, levels = c("IBD", "non-IBD"))
+ return(joined)
+ }
+ }
+
+ #if nothing usable, just return NAs
+ out
+}
+
+#evaluate V1..V3 scores with Wilcoxon, PERMANOVA, weighted RF AUROC
+eval_scores <- function(scores_df) {
+ out <- list()
+ if (!("Group" %in% names(scores_df))) return(out)
+
+ #Wilcoxon PC1/PC2
+ res_w1 <- try(wilcox.test(scores_df$V1 ~ scores_df$Group, exact = FALSE), silent = TRUE)
+ res_w2 <- try(wilcox.test(scores_df$V2 ~ scores_df$Group, exact = FALSE), silent = TRUE)
+ out$wilcox_PC1_p <- if (!inherits(res_w1, "try-error")) res_w1$p.value else NA_real_
+ out$wilcox_PC2_p <- if (!inherits(res_w2, "try-error")) res_w2$p.value else NA_real_
+
+ #PERMANOVA on V1..V3
+ if (all(c("V1", "V2", "V3") %in% names(scores_df))) {
+ perm_df <- na.omit(scores_df[, c("Group", "V1", "V2", "V3")])
+ if (is.factor(perm_df$Group) && nlevels(perm_df$Group) >= 2 && all(table(perm_df$Group) >= 5)) {
+ perm <- vegan::adonis2(perm_df[, c("V1","V2","V3")] ~ Group, data = perm_df, method = "euclidean")
+ out$permanova_R2 <- perm$R2[1]; out$permanova_F <- perm$F[1]; out$permanova_p <- perm$`Pr(>F)`[1]
+ }
+ }
+
+ # Weighted RF AUROC
+ if (all(c("V1", "V2", "V3") %in% names(scores_df))) {
+ rf_df <- na.omit(scores_df[, c("Group", "V1", "V2", "V3")])
+ if (is.factor(rf_df$Group) && nlevels(rf_df$Group) >= 2) {
+ cls_tab <- table(rf_df$Group); wts <- as.numeric(1/cls_tab); names(wts) <- names(cls_tab)
+ set.seed(42)
+ rf_prob <- ranger(Group ~ ., data = rf_df, num.trees = 1000,
+ probability = TRUE, class.weights = wts, oob.error = TRUE)
+ if ("IBD" %in% colnames(rf_prob$predictions)) {
+ p_ibd <- rf_prob$predictions[, "IBD"]
+ roc_obj <- pROC::roc(rf_df$Group, p_ibd, levels = c("non-IBD", "IBD"))
+ out$AUROC <- as.numeric(pROC::auc(roc_obj))
+ }
+ }
+ }
+
+ out
+}
+
+# Load MGX / MTX / VRX and build matrices
+
+dt_mgx <- read_ibdmdb_tsv(f_mgx)
+dt_mtx <- read_ibdmdb_tsv(f_mtx)
+
+cat("\nDims (rows, cols):\n")
+
+
+Dims (rows, cols):cat("MGX:", dim(dt_mgx), "\n")
+
+MGX: 1479 1639 cat("MTX:", dim(dt_mtx), "\n")
+
+MTX: 70711 736 mat_mgx <- to_matrix(dt_mgx)
+mat_mtx <- to_matrix(dt_mtx)
+
+#harmonize samples
+shared_2 <- intersect(colnames(mat_mgx), colnames(mat_mtx))
+shared_2 <- unique(shared_2)
+shared_2 <- shared_2[nchar(shared_2) > 0]
+cat("\nShared MGX–MTX samples:", length(shared_2), "\n")
+
+
+Shared MGX–MTX samples: 734 stopifnot(length(shared_2) >= 20) #sanity threshold
+
+shared_2 <- sort(shared_2)
+X_mgx <- mat_mgx[, shared_2, drop = FALSE]
+X_mtx <- mat_mtx[, shared_2, drop = FALSE]
+
+#guard against duplicate or empty sample IDs
+stopifnot(!any(is.na(shared_2)), all(nchar(shared_2) > 0))
+if (any(duplicated(shared_2))) {
+ message("[guard] Duplicated sample IDs detected; making them unique.")
+ new_ids <- make.unique(shared_2, sep = "_dup")
+ colnames(X_mgx) <- new_ids
+ colnames(X_mtx) <- new_ids
+ shared_2 <- new_ids
+}
+
+# Per-modality adaptive filtering
+
+n_samp <- length(shared_2)
+prev_mgx <- ceiling(0.05 * n_samp) #5% prevalence
+prev_mtx <- ceiling(0.02 * n_samp) #2% prevalence
+
+keep_mgx <- rowSums(X_mgx > 0) >= prev_mgx
+keep_mtx <- rowSums(X_mtx > 0) >= prev_mtx
+
+X_mgx <- X_mgx[keep_mgx, , drop = FALSE]
+X_mtx <- X_mtx[keep_mtx, , drop = FALSE]
+
+#drop samples that are all-zero in a given view
+keep_samp_mgx <- colSums(X_mgx) > 0
+keep_samp_mtx <- colSums(X_mtx) > 0
+if (!all(keep_samp_mgx)) message(sprintf("[MGX] dropping %d all-zero samples", sum(!keep_samp_mgx)))
+if (!all(keep_samp_mtx)) message(sprintf("[MTX] dropping %d all-zero samples", sum(!keep_samp_mtx)))
+X_mgx <- X_mgx[, keep_samp_mgx, drop = FALSE]
+X_mtx <- X_mtx[, keep_samp_mtx, drop = FALSE]
+
+#recompute shared samples after filtering & pruning
+shared_final <- intersect(colnames(X_mgx), colnames(X_mtx))
+shared_final <- sort(unique(shared_final))
+stopifnot(length(shared_final) >= 20)
+
+#subset both views to the same (final) sample set and identical order
+X_mgx <- X_mgx[, shared_final, drop = FALSE]
+X_mtx <- X_mtx[, shared_final, drop = FALSE]
+
+#final sanity on dimensions
+if (nrow(X_mgx) == 0L || nrow(X_mtx) == 0L)
+ stop("After filtering, one view has zero features: MGX rows=", nrow(X_mgx), ", MTX rows=", nrow(X_mtx))
+if (ncol(X_mgx) == 0L || ncol(X_mtx) == 0L)
+ stop("After filtering, there are zero shared samples.")
+
+# Cap MTX by variance for speed
+
+max_mtx_features <- 10000
+if (nrow(X_mtx) > max_mtx_features) {
+mtx_var <- matrixStats::rowVars(X_mtx)
+ord <- order(mtx_var, decreasing = TRUE)
+X_mtx <- X_mtx[ord[seq_len(max_mtx_features)], , drop = FALSE]
+message(sprintf("[MTX] Kept top %d features by variance.", max_mtx_features))
+}
+
+# Clean rownames / numeric
+
+X_mgx <- sanitize_matrix(dedup_rownames(X_mgx))
+X_mtx <- sanitize_matrix(dedup_rownames(X_mtx))
+
+cat("\nRemaining features after filtering:\n",
+"MGX:", nrow(X_mgx), "\n",
+"MTX:", nrow(X_mtx), "\n")
+
+
+Remaining features after filtering:
+ MGX: 459
+ MTX: 10000 cat("Final dims — MGX:", nrow(X_mgx), "x", ncol(X_mgx),
+ "| MTX:", nrow(X_mtx), "x", ncol(X_mtx), "\n")
+
+Final dims — MGX: 459 x 732 | MTX: 10000 x 732 stopifnot(identical(colnames(X_mgx), colnames(X_mtx)))
+
+# Build MAE from the final sample set and run Joint-RPCA
+
+#colData must match the final column set exactly
+cd <- S4Vectors::DataFrame(row.names = colnames(X_mgx))
+
+#SummarizedExperiments (rows = features, cols = samples)
+se_mgx <- SummarizedExperiment::SummarizedExperiment(
+ assays = list(counts = X_mgx),
+ colData = cd
+)
+se_mtx <- SummarizedExperiment::SummarizedExperiment(
+ assays = list(counts = X_mtx),
+ colData = cd
+)
+
+#MultiAssayExperiment
+mae_2 <- MultiAssayExperiment::MultiAssayExperiment(
+ experiments = list(MGX = se_mgx, MTX = se_mtx)
+)
+mae_2 <- MultiAssayExperiment::intersectColumns(mae_2)
+
+#choose a safe rank k
+per_view_min_dim <- sapply(list(X_mgx, X_mtx), function(m) min(nrow(m), ncol(m)))
+k_max <- max(1L, min(per_view_min_dim))
+k <- min(3L, k_max)
+message(sprintf("[2-omic] per-view min dims = %s; using k = %d",
+ paste(per_view_min_dim, collapse = ", "), k))
+
+set.seed(42)
+fit2 <- jointRPCAuniversal(
+ data = mae_2,
+ n.components = k,
+ max.iterations = 500,
+ rclr.transform.tables = TRUE,
+ min.sample.count = 1,
+ min.feature.count = 0,
+ min.feature.frequency = 0
+)
+
+#inspect structure so we don’t assume a specific layout for features
+str(fit2$ord.res, max.level = 2)
+
+List of 7
+ $ method : chr "rpca"
+ $ eigvals : Named num [1:3] 550 489 461
+ ..- attr(*, "names")= chr [1:3] "PC1" "PC2" "PC3"
+ $ samples : num [1:732, 1:3] -0.0397 0.0229 -0.0268 -0.0323 -0.0122 ...
+ ..- attr(*, "dimnames")=List of 2
+ $ features : num [1:10459, 1:3] -0.000878 -0.000878 -0.000878 -0.000878 -0.000878 ...
+ ..- attr(*, "dimnames")=List of 2
+ $ proportion.explained: Named num [1:3] 0.401 0.317 0.282
+ ..- attr(*, "names")= chr [1:3] "PC1" "PC2" "PC3"
+ $ dist : NULL
+ $ metadata : list()
+ - attr(*, "class")= chr "OrdinationResults"#extract sample scores (U)
+U <- as.data.frame(fit2$ord.res$samples)
+colnames(U) <- paste0("V", seq_len(ncol(U)))
+U$sample_id <- rownames(U)
+
+#extract per-modality loadings robustly
+#some builds store ord.res$features as a list per assay; others as a single object
+get_view <- function(obj, keys) {
+ if (is.null(obj)) return(NULL)
+ if (is.list(obj)) {
+ for (k in keys) if (!is.null(obj[[k]])) return(as.data.frame(obj[[k]]))
+ return(NULL)
+ }
+ #fallback: if it's a matrix/data.frame but not split by view, skip splitting
+ if (is.matrix(obj) || is.data.frame(obj)) return(as.data.frame(obj))
+ return(NULL)
+}
+
+V_mgx <- get_view(fit2$ord.res$features, c("MGX","view_MGX", "view_mgx"))
+V_mtx <- get_view(fit2$ord.res$features, c("MTX","view_MTX", "view_mtx"))
+
+#plot ordination (PC1 vs PC2, unlabeled for now)
+print(
+ ggplot(U, aes(V1, V2)) +
+ geom_point(alpha = 0.8) +
+ labs(title = sprintf("Joint-RPCA (MGX + MTX), k = %d — PC1 vs PC2", k),
+ x = "PC1", y = "PC2") +
+ theme_minimal()
+)
+
+
make_groups_autodetect <- function(meta_df, sample_ids, min_frac = 0.01, min_abs = 10L) {
+ #always return a data.frame(sample_id, Group)
+ out <- data.frame(sample_id = sample_ids, Group = factor(NA, levels = c("IBD", "non-IBD")),
+ stringsAsFactors = FALSE)
+ if (is.null(meta_df) || !nrow(meta_df)) return(out)
+
+ md <- as.data.frame(meta_df, stringsAsFactors = FALSE)
+ names(md) <- tolower(trimws(names(md)))
+
+ sid <- tolower(trimws(as.character(sample_ids)))
+ thresh <- max(min_abs, floor(length(sid) * min_frac))
+
+ #overlap of every column with our sample IDs
+ overlaps <- sapply(md, function(col) {
+ x <- tolower(trimws(as.character(col)))
+ sum(!is.na(x) & x %in% sid)
+ })
+
+ #choose the best column, but only if it clears the threshold
+ max_ov <- suppressWarnings(max(overlaps, na.rm = TRUE))
+ if (!is.finite(max_ov) || max_ov < thresh) {
+ message(sprintf("[meta] No metadata column matches sample IDs (max overlap = %s < %d). Group left NA.",
+ as.character(max_ov), thresh))
+ return(out)
+ }
+ best <- names(overlaps)[which.max(overlaps)]
+ message(sprintf("[meta] Chosen sample-id column: '%s' (matches = %d)", best, overlaps[[best]]))
+
+ if (!"diagnosis" %in% names(md)) {
+ message("[meta] 'diagnosis' not found; Group left NA.")
+ return(out)
+ }
+
+ #build IBD / non-IBD from diagnosis
+ dx <- tolower(trimws(as.character(md$diagnosis)))
+ grp <- ifelse(grepl("\\b(uc|cd|ibd)\\b", dx), "IBD",
+ ifelse(grepl("^\\s*non", dx), "non-IBD", NA_character_))
+
+ md$sample_id <- tolower(trimws(as.character(md[[best]])))
+ md$Group <- factor(grp, levels = c("IBD", "non-IBD"))
+
+ join_tbl <- unique(md[, c("sample_id", "Group")])
+ out <- dplyr::left_join(
+ data.frame(sample_id = sid, stringsAsFactors = FALSE),
+ join_tbl,
+ by = "sample_id"
+ )
+ out$sample_id <- sample_ids #restore original casing
+ out
+}
+
+meta_df <- if (file.exists(f_meta)) data.frame(data.table::fread(f_meta)) else NULL
+grp_df <- make_groups_autodetect(meta_df, U$sample_id)
+print(sort(table(grp_df$Group, useNA = "ifany")))
+
+
+non-IBD IBD
+ 187 545 U2 <- dplyr::left_join(U, grp_df, by = "sample_id")
+
+#color the ordination by IBD status
+ggplot(U2, aes(V1, V2, color = Group)) +
+ geom_point(alpha = 0.8, size = 1.2) +
+ labs(title = sprintf("Joint-RPCA (k=%d): PC1 vs PC2 by Group", k)) +
+ theme_minimal()
+
+
#basic stats: Wilcoxon on PC1/PC2, PERMANOVA on PC1–PC3, RF AUROC
+stats <- eval_scores(U2)
+print(stats)
+
+$wilcox_PC1_p
+[1] 0.03047366
+
+$wilcox_PC2_p
+[1] 2.168776e-09
+
+$permanova_R2
+[1] 0.01724885
+
+$permanova_F
+[1] 12.81267
+
+$permanova_p
+[1] 0.001
+
+$AUROC
+[1] 0.7418535#see how many samples couldn’t be labeled
+sum(is.na(U2$Group))
+
+[1] 0#quick top-loading features (only if per-view loadings were available)
+top_k <- 15
+if (!is.null(V_mgx) && ncol(V_mgx) >= 1) {
+ ord <- order(V_mgx[,1], decreasing = TRUE)
+ cat("\nTop MGX features on PC1:\n")
+ print(data.frame(
+ feature = rownames(V_mgx)[ord][seq_len(min(top_k, length(ord)))],
+ loading = V_mgx[ord, 1][seq_len(min(top_k, length(ord)))]
+ ))
+}
+
+
+Top MGX features on PC1:
+ feature
+1 2.7.1.69: Protein-N(pi)-phosphohistidine--sugar phosphotransferase
+2 2.7.1.69: Protein-N(pi)-phosphohistidine--sugar phosphotransferase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+3 2.7.7.6: DNA-directed RNA polymerase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+4 3.6.3.14: H(+)-transporting two-sector ATPase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+5 4.2.1.55: 3-hydroxybutyryl-CoA dehydratase
+6 4.2.1.55: 3-hydroxybutyryl-CoA dehydratase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+7 2.7.7.27: Glucose-1-phosphate adenylyltransferase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+8 1.3.99.2: Transferred entry: 1.3.8.1|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+9 2.7.1.90: Diphosphate--fructose-6-phosphate 1-phosphotransferase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+10 1.3.99.2: Transferred entry: 1.3.8.1
+11 5.3.1.17: 5-dehydro-4-deoxy-D-glucuronate isomerase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+12 1.1.1.35: 3-hydroxyacyl-CoA dehydrogenase
+13 2.3.1.9: Acetyl-CoA C-acetyltransferase
+14 1.1.1.35: 3-hydroxyacyl-CoA dehydrogenase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+15 1.1.1.157: 3-hydroxybutyryl-CoA dehydrogenase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+ loading
+1 0.06607556
+2 0.06158873
+3 0.06150761
+4 0.06008987
+5 0.05708062
+6 0.05622509
+7 0.05539683
+8 0.05405219
+9 0.05397505
+10 0.05396237
+11 0.05355906
+12 0.05308296
+13 0.05289232
+14 0.05265166
+15 0.05265166if (!is.null(V_mtx) && ncol(V_mtx) >= 1) {
+ ord <- order(V_mtx[,1], decreasing = TRUE)
+ cat("\nTop MTX features on PC1:\n")
+ print(data.frame(
+ feature = rownames(V_mtx)[ord][seq_len(min(top_k, length(ord)))],
+ loading = V_mtx[ord, 1][seq_len(min(top_k, length(ord)))]
+ ))
+}
+
+
+Top MTX features on PC1:
+ feature
+1 2.7.1.69: Protein-N(pi)-phosphohistidine--sugar phosphotransferase
+2 2.7.1.69: Protein-N(pi)-phosphohistidine--sugar phosphotransferase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+3 2.7.7.6: DNA-directed RNA polymerase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+4 3.6.3.14: H(+)-transporting two-sector ATPase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+5 4.2.1.55: 3-hydroxybutyryl-CoA dehydratase
+6 4.2.1.55: 3-hydroxybutyryl-CoA dehydratase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+7 2.7.7.27: Glucose-1-phosphate adenylyltransferase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+8 1.3.99.2: Transferred entry: 1.3.8.1|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+9 2.7.1.90: Diphosphate--fructose-6-phosphate 1-phosphotransferase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+10 1.3.99.2: Transferred entry: 1.3.8.1
+11 5.3.1.17: 5-dehydro-4-deoxy-D-glucuronate isomerase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+12 1.1.1.35: 3-hydroxyacyl-CoA dehydrogenase
+13 2.3.1.9: Acetyl-CoA C-acetyltransferase
+14 1.1.1.35: 3-hydroxyacyl-CoA dehydrogenase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+15 1.1.1.157: 3-hydroxybutyryl-CoA dehydrogenase|g__Faecalibacterium.s__Faecalibacterium_prausnitzii
+ loading
+1 0.06607556
+2 0.06158873
+3 0.06150761
+4 0.06008987
+5 0.05708062
+6 0.05622509
+7 0.05539683
+8 0.05405219
+9 0.05397505
+10 0.05396237
+11 0.05355906
+12 0.05308296
+13 0.05289232
+14 0.05265166
+15 0.05265166
