fusemap 1.1.2__tar.gz → 1.1.3__tar.gz

fusemap-1.1.3/CHANGELOG.md

@@ -0,0 +1,46 @@
+# Changelog
+
+## 1.1.3 — local release preparation, 2026-09-11
+
+Reference mapping could raise `NameError: dist_hooks is not defined` during
+validation following the DDP refactor. `map_model` now defaults `dist_hooks` to
+`None`, so ordinary mapping reaches scheduling and checkpoint selection.
+The regression test executes that path with CPU optimizers and a small model,
+including optional synchronization hooks.
+
+- Complete Tutorials 10–12 and their navigation/data entries. Tutorial 10
+  explicitly uses Leiden as a teaching adaptation of the paper's WGCNA panel
+  workflow. Tutorial 11 distinguishes model-derived marker evidence from
+  independent anatomical validation. Tutorial 12 documents data available upon
+  request, replaces its placeholder download, and interprets the saved mixed
+  domains and anatomically unexpected predictions.
+- Keep classic LangChain 0.3 dependencies in the scientific environment. Agent
+  v2 has a separate requirements file and setup instructions for its LangChain
+  1.x controller and registered scientific kernel. Kernel startup now reports
+  bootstrap errors; SQLite checkpoint directories are created on first use.
+- Replace the obsolete requirements freeze (including a self-pin to FuseMap
+  0.0.1) with installation of the checkout. Bound AnnData below 0.12 for Scanpy
+  1.9.3 compatibility; provide `tutorials` and `agent` extras. Mark the Python 3.7
+  environment YAML as historical.
+- Update the manuscript title, retain the earlier preprint citation as such,
+  document method/version differences, and use the installed Python API in
+  quick starts. Add public API documentation and modern build metadata.
+- Version the corrected artifacts as 1.1.3; keep 1.1.2 artifacts for comparison.
+  Build products and Agent runtime state are excluded from version control.
+
+### Scope
+
+The current manuscript is *An agentic AI spatial molecular foundation model of
+the brain*. Its frozen code and analysis archive is identified at
+https://doi.org/10.5281/zenodo.22103552. Current software enhancements and teaching
+adaptations are described in `docs/release_notes.rst`; they do not replace the
+paper's analysis-specific configurations.
+
+Notebook analysis outputs are retained from earlier runs. Modified setup cells
+are cleared; release preparation does not claim fresh execution of all tutorials,
+full-atlas retraining, live LLM-provider validation, or verification of remote
+data downloads. This checkout and its local build are not a PyPI upload.
+
+### Validation
+
+Validation results are recorded in `RELEASE_VALIDATION.md` after the checks run.

fusemap-1.1.3/MANIFEST.in

@@ -0,0 +1,3 @@
+include README.md LICENSE CHANGELOG.md RELEASE_VALIDATION.md pyproject.toml
+include requirements.txt requirements-agent-v2.txt
+recursive-include tests *.py

{fusemap-1.1.2 → fusemap-1.1.3}/PKG-INFO

@@ -1,10 +1,11 @@
 Metadata-Version: 2.4
 Name: fusemap
-Version: 1.1.2
+Version: 1.1.3
 Requires-Python: >=3.9,<3.12
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: scanpy==1.9.3
+Requires-Dist: anndata<0.12,>=0.9
 Requires-Dist: torch==2.0.1
 Requires-Dist: dgl==1.1.1
 Requires-Dist: sparse==0.14.0
@@ -26,25 +27,36 @@ Requires-Dist: langsmith==0.3.43
 Requires-Dist: easydict==1.13
 Requires-Dist: numpy<2,>=1.26
 Requires-Dist: gdown
+Provides-Extra: tutorials
+Requires-Dist: ipykernel; extra == "tutorials"
+Requires-Dist: scikit-misc<0.6,>=0.3; extra == "tutorials"
+Provides-Extra: agent
+Requires-Dist: tavily-python<1,>=0.7; extra == "agent"
+Requires-Dist: duckduckgo-search<9,>=6; extra == "agent"
 Dynamic: description
 Dynamic: description-content-type
 Dynamic: license-file
+Dynamic: provides-extra
 Dynamic: requires-dist
 Dynamic: requires-python
 
 # FuseMap 
-Integrate spatial transcripomics with universal gene, cell, and tissue embeddings.
+Integrate spatial transcriptomics with universal gene, cell, and tissue embeddings.
 
 <p align="center">
-  <img src="/img/framework.png" width="600" >
+  <img src="https://raw.githubusercontent.com/wanglab-broad/FuseMap/main/docs/_static/framework.png" width="600" >
 </p>
 
-For more details, please check out our publication.
+Current manuscript: **An agentic AI spatial molecular foundation model of the brain**.
+See [release notes](CHANGELOG.md) for the distinction between the manuscript workflows
+and the current software, including anchor alignment and Agent v2.
 
 
 ## Manuscript code and data
 ### Reproducibity
-For code and data in the manuscript 'Towards a universal spatial molecular atlas of the mouse brain', please go to [paper_code](paper_code/).
+For manuscript analysis code, see [paper_code](paper_code/). The current manuscript
+describes a frozen code and analysis archive at [Zenodo](https://doi.org/10.5281/zenodo.22103552).
+Use that archive and its run configurations for paper reproduction.
 
 ### Exploratory analysis
 We provide an interactive online database of the [molCCF](http://fusemap.spatial-atlas.net/).
@@ -74,12 +86,33 @@ seaborn
 
 
 ## Installation and Tutorial
+
+Use Python 3.9–3.11 (3.10 recommended). Install this checkout with:
+
+```bash
+python -m pip install .
+```
+
+For notebook dependencies, use `python -m pip install ".[tutorials]"`.
+The PyPI package is installed with `python -m pip install fusemap`; see the
+[release notes](CHANGELOG.md) for this checkout's version.
+
+```python
+import fusemap
+fusemap.integrate("./data", "./output")
+```
+
+The classic Agent uses `python -m pip install ".[agent]"` from this repository.
+Agent v2 uses a separate environment and a registered scientific kernel; follow
+[the Agent setup instructions](agent_setup/README.md#agent-v2).
+
 - Read the FuseMap tutorial [here](https://fusemap.readthedocs.io/en/latest/).
 - FuseMap-Agent set up tutorial [here](https://github.com/wanglab-broad/FuseMap/tree/main/agent_setup).
 
 
 ## Citation
 
-If you find FuseMap useful for your work, please cite our paper: 
+The current manuscript is titled **An agentic AI spatial molecular foundation model of the brain**.
+The earlier public preprint can be cited as:
 
 > Yichun He, Hao Sheng, Hailing Shi, Wendy Xueyi Wang, Zefang Tang, Jia Liu, Xiao Wang. Towards a universal spatial molecular atlas of the mouse brain. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/2024.05.27.594872v1 (2024).

fusemap-1.1.3/README.md

@@ -0,0 +1,76 @@
+# FuseMap 
+Integrate spatial transcriptomics with universal gene, cell, and tissue embeddings.
+
+<p align="center">
+  <img src="https://raw.githubusercontent.com/wanglab-broad/FuseMap/main/docs/_static/framework.png" width="600" >
+</p>
+
+Current manuscript: **An agentic AI spatial molecular foundation model of the brain**.
+See [release notes](CHANGELOG.md) for the distinction between the manuscript workflows
+and the current software, including anchor alignment and Agent v2.
+
+
+## Manuscript code and data
+### Reproducibity
+For manuscript analysis code, see [paper_code](paper_code/). The current manuscript
+describes a frozen code and analysis archive at [Zenodo](https://doi.org/10.5281/zenodo.22103552).
+Use that archive and its run configurations for paper reproduction.
+
+### Exploratory analysis
+We provide an interactive online database of the [molCCF](http://fusemap.spatial-atlas.net/).
+
+## System Requirements
+### Hardware requirements
+`FuseMap` package requires a standard computer with optional GPU to support the in-memory operations.
+
+### Software requirements
+#### OS Requirements
+This package is supported for *Linux*. The package has been tested on the following system:
++ Linux: Ubuntu 20.04
+
+#### Python Dependencies
+`FuseMap` mainly depends on the Python scientific stack.
+
+```
+dgl
+numpy
+scipy
+scikit-learn
+pandas
+pytorch
+scanpy
+seaborn
+```
+
+
+## Installation and Tutorial
+
+Use Python 3.9–3.11 (3.10 recommended). Install this checkout with:
+
+```bash
+python -m pip install .
+```
+
+For notebook dependencies, use `python -m pip install ".[tutorials]"`.
+The PyPI package is installed with `python -m pip install fusemap`; see the
+[release notes](CHANGELOG.md) for this checkout's version.
+
+```python
+import fusemap
+fusemap.integrate("./data", "./output")
+```
+
+The classic Agent uses `python -m pip install ".[agent]"` from this repository.
+Agent v2 uses a separate environment and a registered scientific kernel; follow
+[the Agent setup instructions](agent_setup/README.md#agent-v2).
+
+- Read the FuseMap tutorial [here](https://fusemap.readthedocs.io/en/latest/).
+- FuseMap-Agent set up tutorial [here](https://github.com/wanglab-broad/FuseMap/tree/main/agent_setup).
+
+
+## Citation
+
+The current manuscript is titled **An agentic AI spatial molecular foundation model of the brain**.
+The earlier public preprint can be cited as:
+
+> Yichun He, Hao Sheng, Hailing Shi, Wendy Xueyi Wang, Zefang Tang, Jia Liu, Xiao Wang. Towards a universal spatial molecular atlas of the mouse brain. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/2024.05.27.594872v1 (2024).

fusemap-1.1.3/RELEASE_VALIDATION.md

@@ -0,0 +1,58 @@
+# FuseMap 1.1.3 release validation
+
+Date: 2026-09-11. These checks apply to the local 1.1.3 working tree and builds.
+They do not certify a GitHub, PyPI, Read the Docs, or Zenodo deployment.
+
+| Check | Result |
+|---|---|
+| Mapping validation/checkpoint regression from source | 2 passed (ordinary mapping and optional synchronization hooks) |
+| Same regression against an installed wheel, outside the source tree | 2 passed with Torch 2.0.1+cu117, Scanpy 1.9.3, AnnData 0.11.4, Python 3.10 |
+| Installed import location/version | Confirmed installation under a temporary virtual environment, FuseMap 1.1.3 |
+| Agent v2 dependency resolution | `pip install --dry-run -r requirements-agent-v2.txt` passed against the controller environment |
+| Wheel plus tutorial/Agent extra dependencies | Online pip dry-run resolved successfully against the pinned tutorial environment |
+| Agent v2 scientific kernel smoke check | Registered a temporary kernel; imported FuseMap 1.1.3, retained a variable across calls, saved a plot, and returned a Python error |
+| Notebook validation | All 12 notebooks: valid schema and 115 code cells compile after IPython syntax transformation |
+| Sphinx HTML documentation | Strict offline build (`-W --keep-going`) passed without warnings |
+| Release archives | Source distribution and wheel built; `twine check` passed |
+| Wheel/source consistency | All 24 package Python files match the working tree byte for byte |
+| Patch whitespace | `git diff --check` passed |
+
+## Reproduce the main checks
+
+From the repository root with the appropriate dependencies installed:
+
+```bash
+python -m pytest tests/test_map_validation.py -q
+FUSEMAP_DOCS_OFFLINE=1 python -m sphinx -b html -W --keep-going docs docs/_build/html
+python -m build
+python -m twine check dist/fusemap-1.1.3*
+```
+
+Install the wheel in a separate environment and run a copy of the tests from
+outside the repository to verify the packaged code rather than importing the
+checkout. `dist/SHA256SUMS-1.1.3` records the final archive checksums.
+
+## Environment and scope
+
+The source regression and kernel smoke check used the existing scientific runtime
+(Torch 2.6.0+cu124); the installed-wheel regression separately checked the declared
+Torch 2.0.1 stack. Temporary test environments and kernel registration were created
+under `/tmp`; existing Conda environments were not changed. The tutorial environment
+had an incomplete optional Dask installation, repaired only in the temporary test
+environment before the installed-wheel checks. Existing library deprecation warnings
+remain; neither mapping regression run reported a test failure after setup.
+
+The first online Sphinx build could not retrieve external intersphinx inventories
+in the restricted network. Local notebook links were fixed, and the complete strict
+build passed with only external inventories disabled via `FUSEMAP_DOCS_OFFLINE=1`.
+
+The mapping tests execute CPU optimizers, validation, learning-rate scheduling and
+checkpoint promotion with a small model, replacing graph preparation and loss
+construction. They are regression checks, not a scientific integration benchmark.
+The kernel smoke check makes no LLM-provider calls and does not load the full atlas.
+
+Existing notebook analysis outputs were retained. Edited setup cells were cleared;
+no fresh full-atlas training, end-to-end tutorial rerun, live LLM analysis, or external
+reference/data download verification is claimed. Tutorial 12 follows the manuscript's
+marmoset data availability-on-request statement. The manuscript and frozen local
+validation snapshots were not modified.

{fusemap-1.1.2 → fusemap-1.1.3}/fusemap/__init__.py

@@ -1,3 +1,5 @@
+__version__ = "1.1.3"
+
 from fusemap.training.integrate import *
 from fusemap.training.map import *
 from fusemap.logger import *

{fusemap-1.1.2 → fusemap-1.1.3}/fusemap/training/train_model.py

@@ -1356,6 +1356,7 @@ def map_model(
     dataloader_pretrain_spatial,
     TRAINED_X_NUM,
     flagconfig,
+    dist_hooks=None,
 ):
     loss_atlas_val_best = float("inf")
     patience_counter = 0

{fusemap-1.1.2 → fusemap-1.1.3}/fusemap.egg-info/PKG-INFO

@@ -1,10 +1,11 @@
 Metadata-Version: 2.4
 Name: fusemap
-Version: 1.1.2
+Version: 1.1.3
 Requires-Python: >=3.9,<3.12
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: scanpy==1.9.3
+Requires-Dist: anndata<0.12,>=0.9
 Requires-Dist: torch==2.0.1
 Requires-Dist: dgl==1.1.1
 Requires-Dist: sparse==0.14.0
@@ -26,25 +27,36 @@ Requires-Dist: langsmith==0.3.43
 Requires-Dist: easydict==1.13
 Requires-Dist: numpy<2,>=1.26
 Requires-Dist: gdown
+Provides-Extra: tutorials
+Requires-Dist: ipykernel; extra == "tutorials"
+Requires-Dist: scikit-misc<0.6,>=0.3; extra == "tutorials"
+Provides-Extra: agent
+Requires-Dist: tavily-python<1,>=0.7; extra == "agent"
+Requires-Dist: duckduckgo-search<9,>=6; extra == "agent"
 Dynamic: description
 Dynamic: description-content-type
 Dynamic: license-file
+Dynamic: provides-extra
 Dynamic: requires-dist
 Dynamic: requires-python
 
 # FuseMap 
-Integrate spatial transcripomics with universal gene, cell, and tissue embeddings.
+Integrate spatial transcriptomics with universal gene, cell, and tissue embeddings.
 
 <p align="center">
-  <img src="/img/framework.png" width="600" >
+  <img src="https://raw.githubusercontent.com/wanglab-broad/FuseMap/main/docs/_static/framework.png" width="600" >
 </p>
 
-For more details, please check out our publication.
+Current manuscript: **An agentic AI spatial molecular foundation model of the brain**.
+See [release notes](CHANGELOG.md) for the distinction between the manuscript workflows
+and the current software, including anchor alignment and Agent v2.
 
 
 ## Manuscript code and data
 ### Reproducibity
-For code and data in the manuscript 'Towards a universal spatial molecular atlas of the mouse brain', please go to [paper_code](paper_code/).
+For manuscript analysis code, see [paper_code](paper_code/). The current manuscript
+describes a frozen code and analysis archive at [Zenodo](https://doi.org/10.5281/zenodo.22103552).
+Use that archive and its run configurations for paper reproduction.
 
 ### Exploratory analysis
 We provide an interactive online database of the [molCCF](http://fusemap.spatial-atlas.net/).
@@ -74,12 +86,33 @@ seaborn
 
 
 ## Installation and Tutorial
+
+Use Python 3.9–3.11 (3.10 recommended). Install this checkout with:
+
+```bash
+python -m pip install .
+```
+
+For notebook dependencies, use `python -m pip install ".[tutorials]"`.
+The PyPI package is installed with `python -m pip install fusemap`; see the
+[release notes](CHANGELOG.md) for this checkout's version.
+
+```python
+import fusemap
+fusemap.integrate("./data", "./output")
+```
+
+The classic Agent uses `python -m pip install ".[agent]"` from this repository.
+Agent v2 uses a separate environment and a registered scientific kernel; follow
+[the Agent setup instructions](agent_setup/README.md#agent-v2).
+
 - Read the FuseMap tutorial [here](https://fusemap.readthedocs.io/en/latest/).
 - FuseMap-Agent set up tutorial [here](https://github.com/wanglab-broad/FuseMap/tree/main/agent_setup).
 
 
 ## Citation
 
-If you find FuseMap useful for your work, please cite our paper: 
+The current manuscript is titled **An agentic AI spatial molecular foundation model of the brain**.
+The earlier public preprint can be cited as:
 
 > Yichun He, Hao Sheng, Hailing Shi, Wendy Xueyi Wang, Zefang Tang, Jia Liu, Xiao Wang. Towards a universal spatial molecular atlas of the mouse brain. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/2024.05.27.594872v1 (2024).

{fusemap-1.1.2 → fusemap-1.1.3}/fusemap.egg-info/SOURCES.txt

@@ -1,5 +1,11 @@
+CHANGELOG.md
 LICENSE
+MANIFEST.in
 README.md
+RELEASE_VALIDATION.md
+pyproject.toml
+requirements-agent-v2.txt
+requirements.txt
 setup.py
 fusemap/__init__.py
 fusemap/api.py
@@ -29,4 +35,5 @@ fusemap/training/ddp.py
 fusemap/training/integrate.py
 fusemap/training/map.py
 fusemap/training/sharded.py
-fusemap/training/train_model.py
+fusemap/training/train_model.py
+tests/test_map_validation.py

{fusemap-1.1.2 → fusemap-1.1.3}/fusemap.egg-info/requires.txt

@@ -1,4 +1,5 @@
 scanpy==1.9.3
+anndata<0.12,>=0.9
 torch==2.0.1
 dgl==1.1.1
 sparse==0.14.0
@@ -20,3 +21,11 @@ langsmith==0.3.43
 easydict==1.13
 numpy<2,>=1.26
 gdown
+
+[agent]
+tavily-python<1,>=0.7
+duckduckgo-search<9,>=6
+
+[tutorials]
+ipykernel
+scikit-misc<0.6,>=0.3

fusemap-1.1.3/pyproject.toml

@@ -0,0 +1,3 @@
+[build-system]
+requires = ["setuptools>=68", "wheel"]
+build-backend = "setuptools.build_meta"

fusemap-1.1.3/requirements-agent-v2.txt

@@ -0,0 +1,11 @@
+# Install in a separate Python 3.11 environment, without fusemap/requirements.txt.
+# The scientific stack runs through the registered fusemap-kernel (see docs/agent).
+langchain>=1.4,<2
+langgraph>=1.2,<2
+langchain-openai>=1.6,<2
+langchain-anthropic>=1.7,<2
+langchain-google-genai>=4.4,<5
+langgraph-checkpoint-sqlite>=3.1,<4
+jupyter-client>=8.6,<9
+streamlit>=1.45,<2
+tavily-python>=0.7,<1

fusemap-1.1.3/requirements.txt

@@ -0,0 +1,3 @@
+# Install the current checkout; dependency versions are maintained in setup.py.
+# Agent v2 uses requirements-agent-v2.txt in a separate environment.
+-e .

{fusemap-1.1.2 → fusemap-1.1.3}/setup.py

@@ -5,10 +5,11 @@ with open('README.md','r') as f:
 
 setup(
     name='fusemap',
-    version='1.1.2',
+    version='1.1.3',
     packages=find_packages(include=['fusemap', 'fusemap.*']),
     install_requires=[
         'scanpy==1.9.3',
+        'anndata>=0.9,<0.12',
         'torch==2.0.1',
         'dgl==1.1.1',
         'sparse==0.14.0',
@@ -32,8 +33,12 @@ setup(
         "numpy>=1.26,<2",
         "gdown"
     ],
+    extras_require={
+        'tutorials': ['ipykernel', 'scikit-misc>=0.3,<0.6'],
+        'agent': ['tavily-python>=0.7,<1', 'duckduckgo-search>=6,<9'],
+    },
     python_requires=">=3.9,<3.12",
     long_description=description,
     long_description_content_type='text/markdown',
 
-)
+)

fusemap-1.1.3/tests/test_map_validation.py

@@ -0,0 +1,59 @@
+"""Exercise the mapping validation/checkpoint path without a large atlas or GPU."""
+import importlib
+from types import SimpleNamespace
+
+import pytest
+import torch
+
+
+@pytest.mark.parametrize("with_hooks", [False, True])
+def test_mapping_reaches_validation_and_writes_final_checkpoint(tmp_path, monkeypatch, with_hooks):
+    training = importlib.import_module("fusemap.training.train_model")
+    value = lambda x: SimpleNamespace(value=x)
+    config = SimpleNamespace(
+        optim_kw=value("RMSprop"), learning_rate=value(0.001),
+        lr_factor_pretrain=value(0.5), lr_patience_pretrain=value(2),
+        epochs_run_pretrain=0, epochs_run_final=0, n_epochs=value(2), n_atlas=1,
+        use_llm_gene_embedding="false", verbose=value(False),
+        TRAIN_WITHOUT_EVAL=value(0), patience_limit_final=value(5),
+        lr_limit_final=value(1e-5), save_dir=str(tmp_path),
+        snapshot_path=str(tmp_path / "snapshot.pt"),
+    )
+    monkeypatch.setattr(training, "ModelType", config)
+    (tmp_path / "trained_model").mkdir()
+    model = torch.nn.Module()
+    for name in ("discriminator_single", "discriminator_spatial", "encoder", "decoder", "scrna_seq_adj"):
+        model.add_module(name, torch.nn.Linear(2, 2))
+
+    # Replace only graph preparation and loss construction; execute the actual
+    # mapping loop, optimizers, validation, scheduling, and checkpoint promotion.
+    monkeypatch.setattr(training, "get_data", lambda *a: (None,) * 9)
+    monkeypatch.setattr(training, "compute_dis_loss_map",
+                        lambda m, *a: {"dis": m.discriminator_single.weight.square().sum()})
+    validations = []
+
+    def ae_loss(m, *args):
+        loss = m.encoder.weight.square().sum()
+        if not torch.is_grad_enabled():
+            validations.append(float(loss))
+        return {"loss_all": loss, "loss_AE_all": [loss], "dis_ae": loss}
+
+    monkeypatch.setattr(training, "compute_ae_loss_map", ae_loss)
+    synced = []
+
+    def sync_value(v):
+        synced.append(v)
+        return v
+
+    kwargs = {"dist_hooks": SimpleNamespace(sync_value=sync_value)} if with_hooks else {}
+    training.map_model(
+        model, [None], None, None, torch.device("cpu"), None, None, str(tmp_path),
+        [[torch.ones(2, 2)]], [[torch.ones(2, 2)]], 1,
+        SimpleNamespace(align_anneal=1e10, lambda_disc_single=1), **kwargs,
+    )
+    assert len(validations) == 1
+    assert len(synced) == int(with_hooks)
+    checkpoint = tmp_path / "trained_model" / "FuseMap_map_model_final.pt"
+    state = torch.load(checkpoint, map_location="cpu", weights_only=True)
+    assert torch.equal(state["encoder.weight"], model.encoder.weight)
+    assert not (tmp_path / "trained_model" / "FuseMap_map_model.pt").exists()

fusemap-1.1.2/README.md

@@ -1,51 +0,0 @@
-# FuseMap 
-Integrate spatial transcripomics with universal gene, cell, and tissue embeddings.
-
-<p align="center">
-  <img src="/img/framework.png" width="600" >
-</p>
-
-For more details, please check out our publication.
-
-
-## Manuscript code and data
-### Reproducibity
-For code and data in the manuscript 'Towards a universal spatial molecular atlas of the mouse brain', please go to [paper_code](paper_code/).
-
-### Exploratory analysis
-We provide an interactive online database of the [molCCF](http://fusemap.spatial-atlas.net/).
-
-## System Requirements
-### Hardware requirements
-`FuseMap` package requires a standard computer with optional GPU to support the in-memory operations.
-
-### Software requirements
-#### OS Requirements
-This package is supported for *Linux*. The package has been tested on the following system:
-+ Linux: Ubuntu 20.04
-
-#### Python Dependencies
-`FuseMap` mainly depends on the Python scientific stack.
-
-```
-dgl
-numpy
-scipy
-scikit-learn
-pandas
-pytorch
-scanpy
-seaborn
-```
-
-
-## Installation and Tutorial
-- Read the FuseMap tutorial [here](https://fusemap.readthedocs.io/en/latest/).
-- FuseMap-Agent set up tutorial [here](https://github.com/wanglab-broad/FuseMap/tree/main/agent_setup).
-
-
-## Citation
-
-If you find FuseMap useful for your work, please cite our paper: 
-
-> Yichun He, Hao Sheng, Hailing Shi, Wendy Xueyi Wang, Zefang Tang, Jia Liu, Xiao Wang. Towards a universal spatial molecular atlas of the mouse brain. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/2024.05.27.594872v1 (2024).