scpclm 0.1.0

Hybrid parametric modeling of single-cell gene expression dynamics along pseudotime

scPCLM

scPCLM is a hybrid parametric framework for modeling gene expression dynamics along pseudotime in single-cell RNA-seq data.

It is designed to identify differentiation-associated genes by fitting biologically interpretable hill-shaped and valley-shaped trajectories, while providing robust estimation of temporal features such as turning points.

---

Features

• Supports multiple count distributions:

  • Poisson
  • Negative Binomial (NB)
  • Zero-Inflated Poisson (ZIP)
  • Zero-Inflated Negative Binomial (ZINB)

• Models asymmetric expression patterns:

  • Hill-shaped trajectories
  • Valley-shaped trajectories

• Hybrid optimization strategy:

  • Global search via Differential Evolution
  • Local refinement via L-BFGS-B

• Interpretable parameters:

  • Turning point (t0)
  • Activation and decay rates (k1, k2)
  • Expression magnitude (mu)

• Uncertainty estimation:

  • Fisher information-based confidence intervals

---

Installation

Install from PyPI:

pip install scpclm

Or install from source:

git clone https://github.com/jesseslight9-jpg/scpclm
cd scpclm
pip install -e .

---

Input Data Format

Input data should be a cell-by-gene matrix where:

• The first column is pseudotime
• The remaining columns are gene expression values

Example:

Index	Pseudotime	Gene1	Gene2
1	t1	y11	y12
2	t2	y21	y22

---

Python API Usage

Automatic mode

from scpclm import scpclm_auto

result = scpclm_auto(
    gene_index=100,
    t=pseudotime,
    y1=expression,
    marginal="ZIP",
    iter_num=100,
    save_dir="Results/"
)

---

Hill-only mode

from scpclm import scpclm_hill

result = scpclm_hill(
    gene_index=100,
    t=pseudotime,
    y1=expression,
    marginal="NB"
)

---

Valley-only mode

from scpclm import scpclm_valley

result = scpclm_valley(
    gene_index=100,
    t=pseudotime,
    y1=expression,
    marginal="ZIP"
)

---

Output

The model returns:

• Estimated parameters: mu, k1, k2, t0
• Model selection criterion: AIC
• Confidence intervals
• Fitted expression values
• Visualization plots (.png)

---

Command-line Usage (Optional)

python scripts/run_scPCLM.py \
    --model.iter 100 \
    --model.marginal ZIP \
    --data.dir your_data.csv \
    --model.save_dir Results/

---

Project Structure

scpclm/
    scpclm/      core implementation
    scripts/     command-line scripts
    examples/    demos

---

License

MIT License

---

Citation

If you use this method, please cite:

scPCLM: A hybrid parametric approach for identifying differentiation-associated genes from single-cell RNA-seq data