Implementation of CellPLM

CHANGELOG.md

@@ -6,21 +6,9 @@
 
 ## NEW FUNCTIONALITY
 
-* Added `control_methods/true_labels` component (PR #5).
-
-* Added `methods/logistic_regression` component (PR #5).
-
-* Added `metrics/accuracy` component (PR #5).
 
 ## MAJOR CHANGES
 
-* Updated `api` files (PR #5).
-
-* Updated configs, components and CI to the latest Viash version (PR #8).
-
-## MINOR CHANGES
 
-* Updated `README.md` (PR #5).
 
-## BUGFIXES
 

src/methods/cellplm/config.vsh.yaml

@@ -0,0 +1,60 @@
+__merge__: ../../api/base_method.yaml
+
+# A unique identifier for your component (required).
+# Can contain only lowercase letters or underscores.
+name: cellplm
+# A relatively short label, used when rendering visualisations (required)
+label: CellPLM
+# A one sentence summary of how this method works (required). Used when 
+# rendering summary tables.
+summary: "A foundation model pre-trained with cells as tokens."
+# A multi-line description of how this component works (required). Used
+# when rendering reference documentation.
+description: |
+  CellPLM is a pre-trained language model specifically designed for single-cell analysis that leverages the principles of natural language processing (NLP) to understand and process single-cell gene expression data. 
+references:
+  doi: 
+    - 10.1101/2023.10.03.560734
+links:
+  # URL to the documentation for this method (required).
+  documentation: https://github.com/OmicsML/CellPLM/tree/main/tutorials
+  # URL to the code repository for this method (required).
+  repository: https://github.com/OmicsML/CellPLM
+
+
+info:
+  method_types: [embedding]
+  preferred_normalization: counts
+
+arguments:
+  - name: --model
+    type: string
+    description: String giving the CellPLM model to use
+    choices: ["20231027_85M"]
+    default: "20231027_85M"
+#   - name: --n_hvg
+#     type: integer
+#     default: 3000
+#     description: Number of highly variable genes to use.
+
+resources:
+  - type: python_script
+    path: script.py
+  - path: /src/utils/read_anndata_partial.py
+
+engines:
+  - type: docker
+    image: openproblems/base_pytorch_nvidia:1.0.0
+    # TODO: Try to find working installation of flash attention (flash-attn<1.0.5)
+    setup:
+      - type: python
+        pypi:
+          - gdown
+          - scgpt # Install from PyPI to get dependencies
+          - cellplm
+
+runners:
+  - type: executable
+  - type: nextflow
+    directives:
+      label: [midtime, midmem, midcpu, gpu]

src/methods/cellplm/script.py

@@ -0,0 +1,96 @@
+import sys
+import tempfile
+import scanpy as sc
+import anndata as ad
+import gdown
+import torch
+
+import warnings
+warnings.filterwarnings("ignore")
+from CellPLM.utils import set_seed
+
+import numpy as np
+import anndata as ad
+from CellPLM.pipeline.cell_embedding import CellEmbeddingPipeline
+
+## VIASH START
+# Note: this section is auto-generated by viash at runtime. To edit it, make changes
+# in config.vsh.yaml and then run `viash config inject config.vsh.yaml`.
+
+par = {
+  'input': 'resources_test/.../input.h5ad',
+  'output': 'output.h5ad',
+   "model": "20231027_85M",
+}
+meta = {
+  'name': 'cellplm'
+}
+## VIASH END
+
+sys.path.append(meta["resources_dir"])
+from read_anndata_partial import read_anndata
+
+set_seed(24)
+PRETRAIN_VERSION = par['model']
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+
+print("\n>>> Reading input files...", flush=True)
+print(f"Input H5AD file: '{par['input']}'", flush=True)
+adata = read_anndata(
+    par['input'],
+    X='layers/normalized',
+    obs='obs',
+    var='var',
+    uns='uns'
+)
+
+if adata.uns["dataset_organism"] != "homo_sapiens":
+    raise ValueError(
+        f"CellPLM can only be used with human data "
+        f"(dataset_organism == \"{adata.uns['dataset_organism']}\")"
+    )
+
+print(adata, flush=True)
+
+print('Preprocess data', flush=True)
+# ... preprocessing ...
+
+print('Train model', flush=True)
+# ... train model ...
+
+drive_path = f"https://drive.google.com/drive/folders/1C2fVNEKX3plHnagaTwpuPW5tpwv1up9G?usp=sharing"
+model_dir = tempfile.TemporaryDirectory()
+print(f"Downloading from '{drive_path}'", flush=True)
+gdown.download_folder(drive_path, output=model_dir.name, quiet=True)
+print(f"Model directory: '{model_dir.name}'", flush=True)
+
+pipeline = CellEmbeddingPipeline(pretrain_prefix=PRETRAIN_VERSION, # Specify the pretrain checkpoint to load
+                                 pretrain_directory=model_dir.name)
+
+# DEVICE ='cpu'
+embedding = pipeline.predict(adata, # An AnnData object
+                device=DEVICE) # Specify a gpu or cpu for model inference
+
+embedding = embedding.cpu().numpy()
+
+print('Generate predictions', flush=True)
+# ... generate predictions ...
+
+output = ad.AnnData(
+    obs=adata.obs[[]],
+    var=adata.var[[]],
+    obsm={
+        "X_emb": embedding,
+    },
+    uns={
+        "dataset_id": adata.uns["dataset_id"],
+        "normalization_id": adata.uns["normalization_id"],
+        "method_id": meta["name"],
+    },
+)
+print(output)
+
+output.write_h5ad(par['output'], compression='gzip')
+
+print("\n>>> Cleaning up temporary directories...", flush=True)
+model_dir.cleanup()