v0.1.1

🚀 GENBoostGPU v0.1.1

Genomic Elastic Net Boosting on GPU (GENBoostGPU)
GPU-accelerated elastic net regression with boosting for large-scale epigenomic data analysis.

---

✨ What’s New in v0.1.1

• Core orchestration functions added

  • run_windows_with_dask: batch orchestration of genomic windows across single or multiple GPUs using Dask.
  • run_single_window: single-region boosting elastic net with flexible inputs (preloaded arrays or file paths).

• Boosting Elastic Net updates

  • Added Optuna-based hyperparameter tuning for ElasticNet (alpha, l1_ratio).
  • Added Ridge regression tuning with delayed evaluation.
  • Early stopping based on variance explained stability.
  • Final Ridge refit for improved betas and heritability estimates.

• SNP preprocessing improvements

  • Zero-variance filtering
  • Missing genotype imputation
  • LD clumping (PLINK-like) implemented in CuPy
  • Cis-window filtering with error region support

• Output enhancements

  • Per-window .parquet summary tables
  • Saved betas and heritability estimates per phenotype-window pair
  • Support for reproducibility with saved model parameters

• Examples

  • examples/vmr_test_caudate.py: run on VMR-defined CpG regions.
  • examples/simu_test_100n.py: run on simulated SNP–phenotype datasets.

---

📦 Installation

Available on [PyPI](https://pypi.org/project/genboostgpu/):

pip install genboostgpu

Requires:

• Python ≥ 3.10
• NVIDIA GPU with CUDA 12.x
• RAPIDS cuML / cuDF / CuPy

---

🔧 Quick Usage

from genboostgpu.vmr_runner import run_single_window

result = run_single_window(
    chrom=21,
    start=10_000,
    end=510_000,
    geno_path="data/chr21_subset.bed",
    pheno_path="data/phenotypes.tsv",
    pheno_id="pheno_379",
    outdir="results",
    n_iter=50,
    n_trials=10
)

print(result)

Full Changelog: v0.1.0...v0.1.1