Installation Guide

Hardware Requirements

Statescope is designed to scale from lightweight demo usage to large cohort-level Bayesian analyses. Performance depends strongly on sample size, parallelization strategy, and GPU VRAM availability.

Recommended Configurations

Small-scale data (tutorials, testing, serial jobs)

• Sample size: ~50 samples
• CPU: 1–4 cores
• Memory: 8–16 GB RAM
• GPU: optional (small VRAM sufficient)

---

Medium-scale data (hundreds of samples)

• CPU: 16–32 threads
• Memory: 64–128 GB RAM
• GPU: optional but recommended

---

Large-scale data (≥1,000 samples, full Bayesian optimization)

• CPU: 32–64 threads (Xeon / EPYC class, ≥2.6 GHz)
• Memory: 128 GB RAM or more
• GPU (recommended):
  • NVIDIA A100 / A40 / RTX 3090 / H100
  • ≥40 GB VRAM strongly recommended

GPU acceleration typically provides a 6–7× speed-up compared to CPU-only execution
(e.g. 3 minutes on GPU vs ~20 minutes on CPU).

Operating System Requirements

Statescope is primarily tested on Linux-based systems:

• Ubuntu 20.04 / 22.04 (recommended)
• CentOS 7 / Rocky Linux 8+

macOS and Windows (via WSL) may work but are not officially benchmarked.

---

Installation

Python Installation

Python Installation

Statescope can be installed directly via pip.
A Conda environment is not required, but can be used if desired.

pip install Statescope

Requirements

• Python ≥ 3.9
• pip < 23.24 (tested with pip 23.23)

R Installation (StatescopeR)

R Installation (Bioconductor)

The R implementation StatescopeR is a separate package maintained independently from the Python version.

• GitHub repository https://github.com/tgac-vumc/StatescopeR

StatescopeR is distributed via Bioconductor (devel) and manages its own runtime dependencies internally using Bioconductor + basilisk.

No Conda environment is required for R usage.

---

Documentation

• https://bioc.r-universe.dev/articles/StatescopeR/StatescopeR.html
• https://bioconductor.org/packages/devel/bioc/html/StatescopeR.html

---

Installation

if (!requireNamespace("BiocManager", quietly = TRUE))
  install.packages("BiocManager")

# Initialize Bioconductor devel
BiocManager::install(version = "devel")

BiocManager::install("StatescopeR")

---

Package Details

• Package: StatescopeR
• Version: 0.99.33
• Bioconductor: 3.23
• R ≥ 4.6.0
• License: MIT

---

Environment Handling (R)

• Python dependencies (if required) are managed internally via basilisk
• Designed for reproducible R ↔ Python workflows using reticulate
• No external Conda or virtualenv setup is needed

GitHub / Conda Environment

GitHub Installation

Clone the Python repository:

git clone https://github.com/tgac-vumc/Statescope.git
cd Statescope

---

Optional: Conda Environment

An environment.yaml file is provided for users who want a fully reproducible setup (Python + optional R tooling).

conda env create -f environment.yaml
conda activate statescope

Environment Includes

• Python dependencies
• Optional R ≥ 4.6
• StatescopeR
• reticulate integration

---

Running a Demo Script

A basic demo notebook is provided in the tutorial folder.

After installation, launch:

jupyter notebook tutorial/BasicTutorial.ipynb

---

Optional: Installing Conda

If you choose to use Conda and do not have it installed:

1. Visit https://docs.conda.io/en/latest/miniconda.html
2. Download the installer for your OS
3. Follow the installation instructions

Verify installation:

conda --version