Working docker | conda configuration found + better parametrization + small changes

Author: gregothebyteknight

conf/modules.config

@@ -409,7 +409,14 @@ process {
         ]
     }
 
-    withName: EXPIMAP {
+    withName: SCARCHES_EXPIMAP {
+        ext.hidden_layer_sizes    = { params.expimap_hidden_layer_sizes }
+        ext.recon_loss            = { params.expimap_recon_loss }
+        ext.n_epochs              = { params.expimap_n_epochs }
+        ext.alpha_epoch_anneal    = { params.expimap_alpha_epoch_anneal }
+        ext.alpha                 = { params.expimap_alpha }
+        ext.alpha_kl              = { params.expimap_alpha_kl }
+        ext.use_early_stopping    = { params.expimap_use_early_stopping }
         publishDir = [
             path: { "${params.outdir}/combine/integrate/expimap" },
             mode: params.publish_dir_mode,

modules/local/scarches/expimap/environment.yml

@@ -1,8 +1,21 @@
 channels:
-  - conda-forge
-  - bioconda
+- conda-forge
+- bioconda
 dependencies:
-  - conda-forge::python=3.14
-  - pip
-  - pip:
-    - scArches=0.6.1
+- conda-forge::pandas=1.5.3
+- conda-forge::python=3.10.0
+- pip
+- pip:
+  - anndata==0.9.2
+  - flax==0.6.11
+  - jax==0.4.17
+  - jaxlib==0.4.17
+  - numpy==1.24.4
+  - pytorch-lightning==1.9.5
+  - scArches==0.6.1
+  - scanpy==1.9.8
+  - scipy==1.10.1
+  - scvi-tools==0.20.3
+  - setuptools==75.3.2
+  - torch==2.4.1
+  - torchmetrics==1.5.2

modules/local/scarches/expimap/main.nf

@@ -4,8 +4,8 @@ process SCARCHES_EXPIMAP {
 
     conda "${moduleDir}/environment.yml"
     container "${workflow.containerEngine == 'singularity' && !task.ext.singularity_pull_docker_container
-            ? 'https://wave.seqera.io/view/builds/bd-e532922f69f9a648_1'
-            : 'community.wave.seqera.io/library/pip_scarches:7e8c7e577326f6ad'}"
+            ? 'https://wave.seqera.io/view/builds/bd-c86ba663882352b6_1'
+            : 'community.wave.seqera.io/library/pandas_python_pip_anndata_pruned:c86ba663882352b6'}"
 
     input:
     tuple val(meta), path(h5ad, arity: 1)
@@ -20,6 +20,13 @@ process SCARCHES_EXPIMAP {
 
     script:
     prefix = task.ext.prefix ?: "${meta.id}"
+    hidden_layer_sizes = task.ext.hidden_layer_sizes ?: [256, 256, 256]
+    recon_loss = task.ext.recon_loss ?: "nb"
+    n_epochs = task.ext.n_epochs ?: 400
+    alpha_epoch_anneal = task.ext.alpha_epoch_anneal ?: 100
+    alpha = task.ext.alpha ?: 0.7
+    alpha_kl = task.ext.alpha_kl ?: 0.5
+    use_early_stopping = task.ext.use_early_stopping ?: true
     if ("${prefix}.h5ad" == "${h5ad}") {
         error "Input and output names are the same, use \"task.ext.prefix\" to disambiguate!"
     }

modules/local/scarches/expimap/templates/expimap.py

@@ -23,6 +23,11 @@
 if "${counts_layer}" != "X":
     adata_processing.X = adata.layers["${counts_layer}"]
 
+# Ensure condition column exists
+condition_col = "${condition_col}"
+if condition_col not in adata_processing.obs.columns:
+    adata_processing.obs[condition_col] = "condition"
+
 # Prior biological knowledge in form of gene programs
 if "${reference_model}":
     sca.utils.add_annotations(adata_processing, "${reference_model}", min_genes=12, clean=True)
@@ -33,17 +38,18 @@
 intr_cvae = sca.models.EXPIMAP(
     adata=adata_processing,
     condition_key="${condition_col}",
-    hidden_layer_sizes=[256, 256, 256],
-    recon_loss="nb"
+    hidden_layer_sizes=${hidden_layer_sizes},
+    recon_loss="${recon_loss}"
 )
 
 # Train the model
+use_early_stopping = "${use_early_stopping}".lower() == 'true'
 intr_cvae.train(
-    n_epochs=400,
-    alpha_epoch_anneal=100,
-    alpha=0.7,
-    alpha_kl=0.5,
-    use_early_stopping=True
+    n_epochs=${n_epochs},
+    alpha_epoch_anneal=${alpha_epoch_anneal},
+    alpha=${alpha},
+    alpha_kl=${alpha_kl},
+    use_early_stopping=use_early_stopping
 )
 
 # Extract the interpretable latent representation

modules/local/scarches/expimap/tests/main.nf.test.snap

@@ -37,5 +37,19 @@
             "nf-test": "0.9.4",
             "nextflow": "25.10.4"
         }
+    },
+    "Should run without failures": {
+        "content": [
+            36984802,
+            24776608,
+            [
+                "versions.yml:md5,b70dc00692171c2b140da84cca2e8be2"
+            ]
+        ],
+        "timestamp": "2026-03-17T01:26:54.147983034",
+        "meta": {
+            "nf-test": "0.9.4",
+            "nextflow": "25.10.4"
+        }
     }
 }

nextflow.config

@@ -73,6 +73,15 @@ params {
     scvi_categorical_covariates   = ''
     scvi_continuous_covariates    = ''
 
+    // Expimap options
+    expimap_hidden_layer_sizes    = [256, 256, 256]
+    expimap_recon_loss            = "nb"
+    expimap_n_epochs              = 400
+    expimap_alpha_epoch_anneal    = 100
+    expimap_alpha                 = 0.7
+    expimap_alpha_kl              = 0.5
+    expimap_use_early_stopping    = true
+
     // Pseudobulking options
     pseudobulk_groupby_labels     = 'batch'
     pseudobulk_min_num_cells      = 5

nextflow_schema.json

@@ -383,6 +383,54 @@
                 }
             }
         },
+        "expimap_options": {
+            "title": "EXPIMAP options",
+            "type": "object",
+            "fa_icon": "fas fa-microscope",
+            "description": "Options for the EXPIMAP integration method",
+            "properties": {
+                "expimap_hidden_layer_sizes": {
+                    "type": "array",
+                    "items": {
+                        "type": "integer"
+                    },
+                    "default": [256, 256, 256],
+                    "description": "Hidden layer sizes for EXPIMAP neural network"
+                },
+                "expimap_recon_loss": {
+                    "type": "string",
+                    "default": "nb",
+                    "enum": ["nb", "zinb", "poisson", "normal"],
+                    "description": "Reconstruction loss for EXPIMAP",
+                    "help_text": "Reconstruction loss for EXPIMAP. Can be 'nb', 'zinb', 'poisson', or 'normal'."
+                },
+                "expimap_n_epochs": {
+                    "type": "integer",
+                    "default": 400,
+                    "description": "Number of epochs for training EXPIMAP"
+                },
+                "expimap_alpha_epoch_anneal": {
+                    "type": "integer",
+                    "default": 100,
+                    "description": "Epoch for alpha annealing in EXPIMAP"
+                },
+                "expimap_alpha": {
+                    "type": "number",
+                    "default": 0.7,
+                    "description": "Alpha parameter for EXPIMAP"
+                },
+                "expimap_alpha_kl": {
+                    "type": "number",
+                    "default": 0.5,
+                    "description": "KL divergence alpha parameter for EXPIMAP"
+                },
+                "expimap_use_early_stopping": {
+                    "type": "boolean",
+                    "default": true,
+                    "description": "Whether to use early stopping for EXPIMAP training"
+                }
+            }
+        },
         "pseudobulking_options": {
             "title": "Pseudobulking options",
             "type": "object",
@@ -593,6 +641,9 @@
         {
             "$ref": "#/$defs/scVI_options"
         },
+        {
+            "$ref": "#/$defs/expimap_options"
+        },
         {
             "$ref": "#/$defs/pseudobulking_options"
         },

tests/nextflow.config

@@ -23,4 +23,9 @@ process {
     withName: CELLBENDER_REMOVEBACKGROUND {
         ext.args = '--epochs 2'
     }
+    withName: SCARCHES_EXPIMAP {
+        ext.n_epochs = 1
+        ext.hidden_layer_sizes = [16, 16]
+        ext.alpha_epoch_anneal = 1
+    }
 }