topic3d

Python package for the decomposition and analysis of chromatin tracing data using topic modeling

Note

This project is under active development. The API is not yet stable and will change frequently.

---

Overview

topic3d enables the analysis and interpretation of 3D chromatin tracing data using Latent Dirichlet Allocation (LDA), which captures recurrent Chromatin Folding Motifs (CFMs) based on pairwise proximity maps.

---

Installation

The package is currently in private development and must be installed from source.

git clone https://github.com/ggautham07/topic3d.git
cd topic3d
pip install .

Dependencies include:
numpy, scipy, matplotlib, seaborn, pandas, scikit-learn, umap-learn, hdbscan, joblib, h5py

---

Usage

To run the full pipeline (train an LDA model and decompose data) on a compatible Fish Omics Format for Chromatin Tracing (FOF-CT) dataset:

topic3d runall \
    --data_train_file path/to/data/file.npy \
    --results_dir path/to/results/folder \
    --proximity_threshold 150 \     # in nm
    --num_CFMs 15 \
    --max_iter 100 \
    --evaluate_learning

---

Authors

• Gautham Ganesh (ggautham07)
• Jean-Bernard Fiche (jbfiche)
• Marcelo Nollmann (marcnol)

---