diff --git a/src/metrics/coranking/config.vsh.yaml b/src/metrics/coranking/config.vsh.yaml
index a269ca6..6984492 100644
--- a/src/metrics/coranking/config.vsh.yaml
+++ b/src/metrics/coranking/config.vsh.yaml
@@ -103,8 +103,9 @@ engines:
     - type: r
       cran:
         - coRanking
+        - Rfast
 runners:
   - type: executable
   - type: nextflow
     directives:
-      label: [hightime, lowmem, lowcpu]
+      label: [hightime, midmem, lowcpu]
diff --git a/src/metrics/coranking/script.R b/src/metrics/coranking/script.R
index a835bdb..170e6e3 100644
--- a/src/metrics/coranking/script.R
+++ b/src/metrics/coranking/script.R
@@ -30,8 +30,8 @@ if (any(is.na(X_emb))) {
   # TODO: computing a square distance matrix is problematic for large datasets!
   # TODO: should we use a different distance metric for the high_dim?
   # TODO: or should we subset to the HVG?
-  dist_highdim <- coRanking:::euclidean(as.matrix(high_dim))
-  dist_emb <- coRanking:::euclidean(as.matrix(X_emb))
+  dist_highdim <- Rfast::Dist(as.matrix(high_dim))
+  dist_emb <- Rfast::Dist(as.matrix(X_emb))
 
   message("Compute ranking matrices")
   rmat_highdim <- rankmatrix(dist_highdim, input = "dist")