3DReconstruction_SpatialTranscriptomics

Overview

This repository banchmarks methods for 3D reconstruction of spatial transcriptomics data on the small MERFISH (Moffit et al. 2018 - "Molecular, spatial, and functional single-cell profiling of the hypothalamic preoptic region") and STARmap datasets (Wang et al. 2018 - "Three-dimensional intact-tissue sequencing of single-cell transcriptional states").

It focuses on reconstructing continuous spatial structures from aligned slices and comparing different computational frameworks for this task (SpatialZ, UOT-based, SpatialZ + PCA/scVI, etc.).

---

Structure

Folders

• data/ – input datasets
• exploration/ – exploratory notebooks and experiments
• inference/ – inference outputs and figures
• output/ – saved results and reconstructions
• uot/ – unbalanced optimal transport (UOT) implementation
• utils/ – helper functions

---

Core Scripts

Reconstruction & methods

• SpatialZ.py – main SpatialZ method

---

Custom methods

• my_method_pca.py – reconstruction using SpatialZ (for cell locations + cell type) and PCA (for gene expression)
• my_method_scvi.py – reconstruction using SpatialZ (for cell locations + cell type) and scVI (for gene expression)
• my_method_pretrained_scvi.py – reconstruction using SpatialZ (for cell locations + cell type) and pre-trained scVI (for gene expression)

---

Comparison scripts

• compare_merfish_small.py – comparison on MERFISH dataset
• compare_starmap.py – comparison on STARmap dataset
• compare_location_models.ipynb – notebook comparing other spatial models (KDE vs. OT)

---

Evaluation

• evaluation_merfish_small.py – evaluation on MERFISH
• evaluation_starmap.py – evaluation on STARmap

---

Data processing & experiments

• cut_dataset.ipynb – dataset preprocessing and slice splitting
• split_train_test*.ipynb – train/test splitting
• optimization_vs_kde.ipynb – experiments comparing other spatial models (KDE vs. OT)

---

Requirements

• requirements_compare.txt – dependencies for comparisons
• requirements_scvi.txt – dependencies for scVI models
• requirements_spatialz.txt – dependencies for SpatialZ