diff --git a/NAMESPACE b/NAMESPACE
index e43ed5d..0823397 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -3,7 +3,6 @@
 export(.extractAMRtable)
 export(.updateBVBRCdata)
 export(CDHIT2duckdb)
-export(buildClusterFeatureMap)
 export(buildDyadFeatureMap)
 export(checkDataAvailability)
 export(cleanData)
diff --git a/R/data_processing.R b/R/data_processing.R
index 4825c16..9f58815 100644
--- a/R/data_processing.R
+++ b/R/data_processing.R
@@ -446,6 +446,7 @@ NULL
     tidyr::separate_rows(protein_ids, sep = ";") |>
     # dplyr::filter(!stringr::str_detect(protein_ids, "_pseudo")) |>
     dplyr::mutate(protein_ids = gsub("_pseudo", "", protein_ids)) |>
+    dplyr::mutate(protein_ids = gsub("_len", "", protein_ids)) |>
     DBI::dbWriteTable(conn = con, name = "genome_gene_protein", overwrite = TRUE)
 }
 
@@ -1405,7 +1406,7 @@ CDHIT2duckdb <- function(duckdb_path,
 #'   Default: `8`.
 #' @param n_workers Integer. Number of parallel HMMER jobs to run. Default: `8`.
 #' @param verbose Logical. Print progress messages. Default: `TRUE`.#'
-#' @returns
+#' @returns Invisibily returns completed HMMER Parquet files per requested database.
 #'
 #' @keywords internal
 .runHMMER <- function(duckdb_path,
@@ -2235,7 +2236,7 @@ CDHIT2duckdb <- function(duckdb_path,
 #' @param ref_file_path Directory containing reference TSVs used by
 #'   [cleanMetaData()] and [cleanData()] for metadata harmonization.
 #'   Default: `"data_raw/"`.
-#' 
+#'
 #' @export
 cleanMetaData <- function(duckdb_path, path, ref_file_path = "data_raw/") {
   duckdb_path <- normalizePath(duckdb_path)
diff --git a/R/feature_to_cluster.R b/R/feature_to_cluster.R
deleted file mode 100644
index 494f52e..0000000
--- a/R/feature_to_cluster.R
+++ /dev/null
@@ -1,171 +0,0 @@
-#' Build a protein cluster-to-feature mapping using DuckDB
-#'
-#' This function constructs a mapping between protein clusters and functional
-#' features (e.g., gene families, domains, COGs, and ARGs) using a DuckDB-backed
-#' workflow. The implementation is SQL-first and memory-efficient, leveraging
-#' DuckDB views and Parquet output.
-#'
-#' @param duckdb_parquet_path Character. Path to the DuckDB database file with parquet file views
-#'   containing the input tables.
-#' @param output_path Character or NULL. Directory where the output Parquet file
-#'   (\code{cluster_feature.parquet}) will be written. If NULL, the output is written
-#'   alongside the DuckDB database.
-#'
-#' @details
-#' The function performs the following steps:
-#' \itemize{
-#'   \item Creates views for each feature type:
-#'     \itemize{
-#'       \item Gene → protein features
-#'       \item Domain annotations
-#'       \item COG annotations
-#'       \item Antibiotic resistance gene (ARG) annotations
-#'     }
-#'   \item Combines all feature mappings into a unified protein–feature view
-#'   \item Joins protein–feature mappings to cluster membership
-#'   \item Writes the resulting cluster–feature mapping to a compressed Parquet file
-#' }
-#'
-#' All joins and transformations are executed inside DuckDB, ensuring scalability
-#' for large datasets without loading data into R memory.
-#'
-#' @return Invisibly returns the file path to the generated Parquet file.
-#'
-#' @examples
-#' \dontrun{
-#' buildClusterFeatureMap(
-#'   duckdb_parquet_path = "data/Csp_parquet.duckdb",
-#'   output_path = NULL
-#' )
-#' }
-#'
-#' @import DBI duckdb
-#' @export
-buildClusterFeatureMap <- function(
-  duckdb_parquet_path,
-  output_path = NULL
-) {
-  con <- DBI::dbConnect(
-    duckdb::duckdb(),
-    normalizePath(duckdb_parquet_path)
-  )
-  on.exit(DBI::dbDisconnect(con, shutdown = TRUE), add = TRUE)
-
-  out_dir <- if (is.null(output_path)) {
-    dirname(duckdb_parquet_path)
-  } else {
-    normalizePath(output_path)
-  }
-  parquet_path <- file.path(out_dir, "cluster_feature.parquet")
-
-  # =========================
-  # Gene → protein features
-  # =========================
-  DBI::dbExecute(con, "
-    CREATE OR REPLACE VIEW v_gene AS
-    SELECT DISTINCT
-      protein_ids AS protein_id,
-      REPLACE(Gene, '~', '.') AS feature
-    FROM genome_gene_protein
-    WHERE protein_ids IS NOT NULL
-      AND Gene IS NOT NULL
-  ")
-
-  # =========================
-  # Domain features
-  # =========================
-  DBI::dbExecute(con, "
-    CREATE OR REPLACE VIEW v_domain AS
-    SELECT DISTINCT
-      AccNum AS protein_id,
-      \"DB.ID\" AS feature
-    FROM domain_names
-    WHERE AccNum IS NOT NULL
-      AND \"DB.ID\" IS NOT NULL
-  ")
-
-  # =========================
-  # Structural gene features
-  # (equivalent to separate_rows + inner_join(gp))
-  # =========================
-  #  DBI::dbExecute(con, "
-  #   CREATE OR REPLACE VIEW v_struct AS
-  #    SELECT DISTINCT
-  #      gp.protein_ids AS protein_id,
-  #     s_gene AS feature
-  #   FROM struct s
-  #   JOIN genome_gene_protein gp
-  #      ON gp.genome_ids = s.genome_id
-  #   CROSS JOIN UNNEST(string_split(s.struct, '.')) AS t(s_gene)
-  #   WHERE s.value = 1
-  #  ")
-
-  # =========================
-  # COG features
-  # =========================
-  DBI::dbExecute(con, "
-    CREATE OR REPLACE VIEW v_cog AS
-    SELECT DISTINCT
-      query_name AS protein_id,
-      name AS feature
-    FROM protein_COG
-    WHERE query_name IS NOT NULL
-      AND name IS NOT NULL
-  ")
-
-  # =========================
-  # ARG features
-  # =========================
-  DBI::dbExecute(con, "
-    CREATE OR REPLACE VIEW v_arg AS
-    SELECT DISTINCT
-      query_name AS protein_id,
-      REPLACE(REPLACE(name, '-NCBIFAM', ''), '-', '.') AS feature
-    FROM protein_ResFinder
-    WHERE query_name IS NOT NULL
-      AND name IS NOT NULL
-  ")
-
-  # =========================
-  # Union: protein → feature
-  # =========================
-  DBI::dbExecute(con, "
-    CREATE OR REPLACE VIEW v_protein_feature AS
-    SELECT protein_id, feature FROM v_gene
-    UNION
-    SELECT protein_id, feature FROM v_domain
-    UNION
-    SELECT protein_id, feature FROM v_cog
-    UNION
-    SELECT protein_id, feature FROM v_arg
-  ")
-
-  # =========================
-  # Cluster → feature mapping
-  # =========================
-  DBI::dbExecute(con, "
-    CREATE OR REPLACE VIEW cluster_feature AS
-    SELECT DISTINCT
-      cm.cluster,
-      pf.feature
-    FROM protein_members cm
-    JOIN v_protein_feature pf
-      ON pf.protein_id = cm.member
-    WHERE cm.cluster IS NOT NULL
-      AND pf.feature IS NOT NULL
-  ")
-
-  # =========================
-  # Write Parquet from DuckDB
-  # =========================
-  DBI::dbExecute(
-    con,
-    sprintf(
-      "COPY cluster_feature TO '%s'
-       (FORMAT PARQUET, COMPRESSION ZSTD)",
-      parquet_path
-    )
-  )
-
-  invisible(parquet_path)
-}
diff --git a/R/feature_to_head.R b/R/feature_to_head.R
index 17bead1..9486096 100644
--- a/R/feature_to_head.R
+++ b/R/feature_to_head.R
@@ -411,7 +411,7 @@ buildDyadFeatureMap <- function(
       view_name = "v_pfam",
       parquet_dir = parquet_dir,
       dataset_name = "protein_Pfam",
-      feature_expr = "query_name"
+      feature_expr = "REPLACE(query_name, '-', '.')"
     )
   }
 
@@ -444,7 +444,7 @@ buildDyadFeatureMap <- function(
       view_name = "v_defensecas",
       parquet_dir = parquet_dir,
       dataset_name = "protein_DefenseCas",
-      feature_expr = "query_name"
+      feature_expr = "REPLACE(query_name, '-', '.')"
     )
   }
 
diff --git a/man/buildClusterFeatureMap.Rd b/man/buildClusterFeatureMap.Rd
deleted file mode 100644
index 7b56363..0000000
--- a/man/buildClusterFeatureMap.Rd
+++ /dev/null
@@ -1,52 +0,0 @@
-% Generated by roxygen2: do not edit by hand
-% Please edit documentation in R/feature_to_cluster.R
-\name{buildClusterFeatureMap}
-\alias{buildClusterFeatureMap}
-\title{Build a protein cluster-to-feature mapping using DuckDB}
-\usage{
-buildClusterFeatureMap(duckdb_parquet_path, output_path = NULL)
-}
-\arguments{
-\item{duckdb_parquet_path}{Character. Path to the DuckDB database file with parquet file views
-containing the input tables.}
-
-\item{output_path}{Character or NULL. Directory where the output Parquet file
-(\code{cluster_feature.parquet}) will be written. If NULL, the output is written
-alongside the DuckDB database.}
-}
-\value{
-Invisibly returns the file path to the generated Parquet file.
-}
-\description{
-This function constructs a mapping between protein clusters and functional
-features (e.g., gene families, domains, COGs, and ARGs) using a DuckDB-backed
-workflow. The implementation is SQL-first and memory-efficient, leveraging
-DuckDB views and Parquet output.
-}
-\details{
-The function performs the following steps:
-\itemize{
-\item Creates views for each feature type:
-\itemize{
-\item Gene → protein features
-\item Domain annotations
-\item COG annotations
-\item Antibiotic resistance gene (ARG) annotations
-}
-\item Combines all feature mappings into a unified protein–feature view
-\item Joins protein–feature mappings to cluster membership
-\item Writes the resulting cluster–feature mapping to a compressed Parquet file
-}
-
-All joins and transformations are executed inside DuckDB, ensuring scalability
-for large datasets without loading data into R memory.
-}
-\examples{
-\dontrun{
-buildClusterFeatureMap(
-  duckdb_parquet_path = "data/Csp_parquet.duckdb",
-  output_path = NULL
-)
-}
-
-}
diff --git a/man/class_abbr.Rd b/man/class_abbr.Rd
index 7863770..dda73de 100644
--- a/man/class_abbr.Rd
+++ b/man/class_abbr.Rd
@@ -11,7 +11,7 @@ A data frame
 Internal reference data
 }
 \usage{
-class_abbr
+data(class_abbr)
 }
 \description{
 A dataset mapping drug classes to their abbreviations.
diff --git a/man/clean_drug.Rd b/man/clean_drug.Rd
index fc4b485..809676c 100644
--- a/man/clean_drug.Rd
+++ b/man/clean_drug.Rd
@@ -11,7 +11,7 @@ A data frame
 Internal reference data
 }
 \usage{
-clean_drug
+data(clean_drug)
 }
 \description{
 A dataset mapping original drug names to cleaned/standardized names.
diff --git a/man/cleaned_bvbrc_countries.Rd b/man/cleaned_bvbrc_countries.Rd
index 4edf9aa..dfda707 100644
--- a/man/cleaned_bvbrc_countries.Rd
+++ b/man/cleaned_bvbrc_countries.Rd
@@ -11,7 +11,7 @@ A data frame
 Internal reference data
 }
 \usage{
-cleaned_bvbrc_countries
+data(cleaned_bvbrc_countries)
 }
 \description{
 A dataset mapping raw country entries to cleaned and standardized names.
diff --git a/man/dot-runHMMER.Rd b/man/dot-runHMMER.Rd
index d693531..e950621 100644
--- a/man/dot-runHMMER.Rd
+++ b/man/dot-runHMMER.Rd
@@ -42,6 +42,9 @@ Default: \code{8}.}
 
 \item{verbose}{Logical. Print progress messages. Default: \code{TRUE}.#'}
 }
+\value{
+Invisibily returns completed HMMER Parquet files per requested database.
+}
 \description{
 Wrapper for preparing HMM databases and running HMMER on protein sequences from duckdb and writing them.
 }
diff --git a/man/drug_abbr.Rd b/man/drug_abbr.Rd
index 39542f2..6912d4f 100644
--- a/man/drug_abbr.Rd
+++ b/man/drug_abbr.Rd
@@ -11,7 +11,7 @@ A data frame
 Internal reference data
 }
 \usage{
-drug_abbr
+data(drug_abbr)
 }
 \description{
 A dataset mapping drug names to their abbreviations.
diff --git a/man/drug_class.Rd b/man/drug_class.Rd
index 1cf412a..89e354a 100644
--- a/man/drug_class.Rd
+++ b/man/drug_class.Rd
@@ -11,7 +11,7 @@ A data frame
 Internal reference data
 }
 \usage{
-drug_class
+data(drug_class)
 }
 \description{
 A dataset mapping drugs to their drug classes.
diff --git a/vignettes/BVBRC_stats.Rmd b/vignettes/BVBRC_stats.Rmd
index cd6e380..b8c27ea 100644
--- a/vignettes/BVBRC_stats.Rmd
+++ b/vignettes/BVBRC_stats.Rmd
@@ -298,6 +298,27 @@ clean_all_char_cols <- function(df) {
     ))
 }
 
+# Clean ALL List columns in the data frame
+clean_list_col <- function(x) {
+  purrr::map(
+    x,
+    \(z) {
+      if (is.null(z) || length(z) == 0)
+        return(NA_character_)
+
+      z <- as.character(unlist(z))
+      z <- stringr::str_squish(z)
+
+      z[is_placeholder_na(z)] <- NA_character_
+
+      if (all(is.na(z)))
+        return(NA_character_)
+
+      z
+    }
+  )
+}
+
 # --- Block-level summaries ---------------------------------------------------
 summarize_block <- function(df, cols, block_name) {
   if (length(cols) == 0) {
@@ -337,6 +358,40 @@ bvbrc <- fetchCompleteBVBRCMetadataAPI()
 
 # Apply to your table
 bvbrc_clean <- clean_all_char_cols(bvbrc)
+list_cols <- names(bvbrc_clean)[sapply(bvbrc_clean, is.list)]
+bvbrc_clean <- bvbrc_clean |>
+  mutate(
+    across(all_of(list_cols), clean_list_col)
+  )
+
+# Save the table to parquet
+
+flatten_list <- function(x) {
+  purrr::map_chr(
+    x,
+    \(z) {
+      if (is.null(z) || length(z) == 0)
+        return(NA_character_)
+
+      paste(as.character(unlist(z)), collapse = ";")
+    }
+  )
+}
+
+bvbrc_clean <- bvbrc_clean |>
+  dplyr::mutate(
+    dplyr::across(
+      dplyr::all_of(list_cols),
+      flatten_list
+    )
+  )
+
+arrow::write_parquet(
+  bvbrc_clean,
+  "data/bvbrc_clean.parquet",
+  compression = "zstd"
+)
+
 # --- Per-column stats --------------------------------------------------------
 col_stats <- tibble::tibble(column = colnames(bvbrc_clean)) |>
   dplyr::mutate(